AI-Art Mastery: Midjourney, Dall-E, Adobe Firefly & Stable Diffusion

1. Welcome: AI will not take your job, but someone that knows how to use AI might will. This comes from Richard Baldwin. Are you ready to be part of this revolution, understand the basics, but also create your own AI models and AI apps? Then this course is for you. And by the way, if you ask yourself who I am, my name is Arne, and I teach AI classes before CGPD even was a thing. So I am relatively long in the game. I also have a small German YouTube channel, and that's what I do. This course is for everything and for every one, regardless of your AI skill set, you don't need to be a programmer because we do every single thing step by step from the basics to the most advanced concepts. All the details about difusion models like M churn, Adobe a fly, stable difusion and Dali to create stunning images. Advanced techniques like fine tuning LMS and Laura's, with which we can even create images of ourselves. The sooner you sign in, the greater the chance to be ahead of competition. So don't test dedate. Click the button now because your future starts today and not tomorrow. 2. The Diffusion Model explained in simple terms: In this video, we will look at what diffusion models are and what they do. So I have found a really, really nice article from medium. This article is relatively long, but we won't make it that far. All I need is this picture right here. Let's assume we have a big big computer and we train our computer on images on images like this. We give the computer images, for example of this beach and we describe it with a text. We give the computer the image, and we say maybe a beach with the blue ocean blue sky. There's some green on the mountains and so on, we are really, really specific. After that, we add some noise to the picture, like you see here, but we still describe what's on the picture. So a beach, blue ocean, blue sky, and so on. More noise, same text, more noise, same text, more noise, same text until you get only noise. In this process, the computer learns how these pictures look like. In this process, he simply understands that the words that you gave the computer yield to this picture. So we can reverse this. If we have only noise and we tell the computer a beach, blue sky, blue ocean. There's some green on the mountains, and so on, the computer can reverse this and make out of the noise this picture. This is really, really a cool concept. And of course, we don't do this with just one picture. We try to give the computer every picture that we can find, and there are, of course, different difusion models. For example, there's also Adobe firefly. Adobe firefly is trained on pictures of Adobe stock. Stable diffusion is open source, and it's free. Everybody can use it, and stable diffusion was trained on pictures from the Internet. And because of this, we also can create nearly everything that is on the Internet. We can create even celebrities. We can create not safe for work stuff and so on. Stable diffusion is not restricted. Nearly everything that is in the Internet, we can create with stable difusion if we give the right prompts. The prompts are the descriptions that we give the computer to make our picture. And for that instance, it's really, really important to make good prompts because we need good pictures. If we are not specific, we can create a picture that look like this. If we simply tell maybe a beach, we will get a random beach. If we tell him a beach, blue ocean, blue, sky, and so on, we will get exactly this picture. So a quick illustration of this process because some people like this illustration, I use this a lot. Just imagine you lay down on the ground and you look in the sky. Besides you, is your girlfriend or your boyfriend or whoever you want. And she tells to you. Can you see this cloud? It looks a little bit like an apple. But you don't get it. You don't see the apple. But then she tells you. Of course, just look. Here is the apple. And then you start to understand you see the cloud and now your eyes see an apple because your brain is trained on apples. Your brain most likely knows how our apple looks like. And then you see the apple in the cloud? Even if there's no apple there. And if your girlfriend doesn't say it's maybe a green apple, maybe you think of a red apple, and that's exactly why we need to use good prompt engineering. Because if we don't are specific, we will get random pictures. If you want to have a green apple, you need to tell the computer that you want to have a green apple. Just like your girlfriend need to tell you that the apple in the clouds is green. If she doesn't tell you that, maybe you'll think of a red apple, maybe of a green apple, maybe even a yellow apple, you don't know so you need to be specific. So in this video, we took a quick look at the diffusion model. The diffusion model works simple. It's trained on pictures and on text. Then noise gets added. The computer learns in this process how this picture looks like. And if we give the computer text afterwards, it can just create these pictures because it will randomly select the pixels that are right for our picture. I hope this makes sense for you. 3. Quick Example of Picture generation in Dall E and Midjourney: You have already learned that there are a lot of different division models. We have stable division, we have mid journey, we have Dali, we have Adobe firefly. And stable division is one of the special division models because with stable division, we can train our own stuff. So in that manner, stable difusion is different, is a bit special. If we just want to use, for example, Dali, you can simply go into the chat PD Abo or you can do it via the copilot. You already know that, and you can simply press on chat PD. If you scroll down and go by made by pa, you can simply press on this difusion model. And here you can simply type in your prompt and you get the picture. Illustration of a cat that plays with a baby dog. And here we have the enormous huge illustrations of a cat that plays with a baby dog. So you see this works really, really nice and really, really good. Dali is one difusion model that is really, really easy to use, and that's why, of course, we look at Dali first in the next section. Then we also have Mid journey, of course, and you already saw Mourne review times. If I take the same prompt and I simply send it up, so don't worry. We will go into detail about Mjourney. You see that Mourney does similar stuff, and I Midourney you can also see how the diffusion model creates the pictures out of the noise. So you can see right here we start with a lot of noise and more noise will get excluded, and then we get really, really good pictures. This takes a bit of time. It depends about how many people are using the division model. Right now we are at 36%, and you can see more and more of the stuff that you will see later on the picture. So the division model simply calculates what pixel needs to be at what place and what color all of these should have. Mi Journey is one of the best difusion models because Marney is trained on really, really large amounts of data and can do really amazing stuff. You see also this picture right here of the CEO of Microsoft. This, for example, is Nadella. This is really a picture that Marney has made. This looks completely like a close up photo of Nadella. The CEO of Microsoft looks exactly like this because Marney is trained on a large amount of data also about celebrities. This is enormous. And this right here are our illustrations. They look really, really cool. I think Dali made the cuter pictures, but Miterny also does a really great job. So we have mid Journey, we have Dali and of course, we have also adobe firefly and stable division. And like I told you, stable division is a bit special because in stable difusion, we have a lot of different models, a lot of different checkpoints, and we have Lauras and we can even train our own checkpoints and auras. And because of this in the next video, I want to show you how this all works, so how stable division is structured. In this section, I want to make a quick deep dive into this theory, and then we start with the easiest the division models with Dali and we see everything about the prompt engineering and so on. But first, I want that you understand the basics. So we need to understand what division models are out there and why they are different. We need to understand this in order to make the right pictures in the right division model. So stay tuned, the next video gets a bit more technical because we simply see what models checkpoints and auras are. And you will also see that we can train them. 4. Stable Diffusion explained: We have learned that stable division is a difusion model. In this video, we will take a closer look and we need to see what as are and what checkpoints are. These two things are really important. We make this really simple in paint. This right here, just think of this as stable difusion. This is the big big computer. This is just a program stable difusion. This is trained, of course, on a lot of pictures. But this enormous diffusion model has, of course, more versions, more checkpoints. And there are different checkpoints for different stuff that we want to make. There is, for example, stable diffusion X L. That's basically just a model that makes really good pictures, maybe from faces, maybe really, really good realistic ones. We will use such a checkpoint. So we have also stable division 1.5. And we have, of course, stable division two point, and so on. Just think about this, there are a lot of different stable division models, stable division checkpoints. And they are always trained on more data. So stable division is the big thing, and then we have the checkpoints. Stable division Excel is, for example, something that makes really, really good realistic pictures, and it depends what you want to make. We have also stuff that makes really really good anime pictures and so on. The important stuff is that the training of this picture takes a lot of time and is not that specific. Here is a big article. We just take a quick look at this. They want to explain to you that there are a lot of different difusion models on stable division, different checkpoints. Stable division 1.41 0.5 realistic version, dream shape. Stable division X L anything version three, deliberate version two. This delibrate for example is perfect for animes. Stable difusion is perfect for realistic pictures. Also this realistic version is good for realistic pictures. The concept here is just that every different checkpoint is trained on a bit of different pictures. So here we have a good example. We have the realistic version, we have the entry, and we have the dreamer. You can see the realistic version makes, of course, realistic pictures. The thing version is a bit more like me, and the dream sharper is also in this direction, and it's realistic painting style. So very different checkpoint is simply a different style. And that's basically a checkpoint in stable division. The model version 1.4 is good in such pictures, and another model is better in other pictures. Generally speaking, we need to use realistic versions, and they are simply trained on realistic pictures to get realistic outputs. I hope all of this makes sense. Then we have the stream sharper that makes good animus and so on. The important stuff is that all of these models, even the dream sharper and so on, these are always really, really big models, and we only have a handful. But we have a really cool solution. We have also auras. There is the possibility to use Laura's and to fine tune these models even further. Think about it this way. Let's say we use stable diffusion, but we want a really specific output. Maybe we want a picture that looks exactly the same in almost every picture. We can fine tune this stable fusion Excel with A Lura this works really nice. We work on this checkpoint, and then we take a little piece of this checkpoint, this piece we call Laura, and this Laura we train on additional pictures, and it's irrelevant what pictures you use. I trained myself, for example, Laura on these pictures. So it's simply my face in different poses. If we train a model on all of these pictures, we can create output that looks similar. But I didn't use stable difusion Excel. I used a model that makes like eima pictures. This is, for example, a picture that I have made with this model. First, we train our model on a lot of pictures of me. The model or the checkpoint was stable diffusion at dream sharper model, and the Laura is then a fine tuned Laura of myself. That's also a picture of the model, and this, like you can see, you can make different kinds of output. And the cool stuff is we don't have to train our auras ourself. The training of the Laura takes a bit of time, but we can search for auras that we like because there are millions of auras out there. And if we take the right Laura and we take it every time, we will get consistent characters. So in this video, we took a closer look at stale division. You have learned that stable division is the big difusion model. Then there are checkpoints. The checkpoints are something like the versions of stable division. They are, of course, trained on different styles. You have a realistic one so stable division Excel. You have the dream sharper for animus and so on, and they make really, really good pictures, but they are really, really big. It takes months of training for such models for such checkpoints. And then we have Laura's. Laura's are smaller checkpoints that we can plug in in our big checkpoints, and with Laura's, we can fine tune our models, our checkpoints. We can fine tune our auras ourself, but we can also take Claas that other people have trained, and we will search perfect auras. That's basically the concept of stable diffusion. In the next video, we take a quick look at the sat, and then we are ready to go. 5. The Seed, the magic Number for consistentPictures in Diffusion Models: In this video, we will talk about sats. The sat is simply the first starting noise of the difusion model. You have learned that all these models make pictures out of the noise. If we give the diffusion model a sat, the sat is a random number, the diffusion model will start from a specific start point. If we give the number ten, for example, the diffusion model will look at the tenth picture where it was trained on and it will generate a similar picture. If we take a random sat, we will get more random pictures. That's the basic of the sat. Now, come with me in C R and take a little look. So we have here the prompt. We have also auras and so on. But first, I want to show you simple what the seed does. We have our prompt and we get the same picture twice, exactly the same picture. If I scroll up a bit, we have the same prompt, but a different picture appears. That's the basic of the seed. If I press generate, here we have the sat. If the sat is minus one, we will take a random seed. If we take the sat, for example, at 20, we will use the twentyith picture to start our generation. If I press sent, now we will get a picture that is different than this picture, but it's the same prompt and all. If I do this again, so simply generate, and I use the same sat, so 20, we will get the same picture twice. And then for proof, I will do it one last time. I press generate. But this time, I will use another seat. We can use a random seat or even just one picture lower and we will get a different output. So now I will send this, and you will see. So this is the first picture. The second should be exactly the same pictures. Yes, you can see this. So we have the same prompt, the same seat, the same lauras and so on, and we get exactly the same pictures. And then we take a seat that is a little bit lower and we will get different pictures. Yes, you can see. We have nearly the same girl because we use auras, we use checkpoints and so on, and we use, of course, the same prompt, but simply the seat does here the magic. We can create exactly the same picture twice. If we use the same seat, the same auras, the same checkpoints and the same prompt. Same prompt, same aura, same checkpoint, but different seat will generate slightly different variations of our picture. If we take same aura, same seat, same checkpoints, but a different prompt, we can create the most consistent characters. I would argue that also this works really, really fine. This are also nearly the same girls, but you can see the seed is here the missing link that some people don't get. In this video, you have learned what the seed is and what the seed does. The seat is simply the first starting noise of the diffusion model. Same sat with same prompt will generate exactly the same picture. Same prompt, but different will generate slightly different variations of this character. Same seat, and a little bit of a different prompt will take our character and we can make our character in different environments. But you have always to check that you use the right auras, the right checkpoints, and so on. This was just the illustration of the sat. 6. Conclusions of the Basics and the Outlook: Until now you have learned that there are a lot of different division models. We have stable division and stable division is the biggest and most complex division model because we have checkpoints, we have models, and we have auras. But stable division is also the most interesting because we can't train our own models. Besides that, we have Murney. Murney is one of the best difusion models. But Murney is not completely for free. Besides stable difusion and Mourney we also have Adobe firefly. Firefly is the difusion model of Adobe. And this is trained on pictures of Adobe stock. This is also special because this is perfect for copyrights. So we have no issues with copyrights ever if we use Adobe firefly. Besides the models, we also have, of course, Dali. Dali is the difusion model of PMI, and we can use this in CCPD and also in their own platform, and also in the copilot. We have a lot of different options to use Dali, and Dali is the difusion model that is the easiest to use. And because of that, we start with Dali over the next section, and just have fun because now it gets interesting because we start to generating pictures. Every division model is a bit different. We need other prompt engineering for every division model, but the concept is always the same. They need the right information, and nearly all of these division models also work with. So stay tuned over the next sections because now we will make our own pictures. And like I said, we start with the easiest model so that we can generate really, really good pictures fast. 7. How you can get access to Dall E [for Free] ALL the ways ChatGPT, Labs, Copilot, Bing, API: The next lectures are all about Dali. Because Dali is the difusion model that is first of all easiest to use, and it makes really, really good and precise pictures. Dali is created from OPM, so the same company that also created CCPD. Dali was also one of the first difusion models. But it all started out in other interfaces and, of course, with much, much worse quality. In this video, I just want to show you where you can get access to Dali because we have a lot of different options, and some of them are totally for free. So just choose the option that is right for you, and I think you can and will create absolute fantastic pictures. So if you have a CCPD pla and I think this will also get rolled out in the free version over time. You can simply press on explore and search the right CPD as you already know. You can simply press on Dali and here you can generate pictures. And that's not all. You can also go on the normal CPD interface right here and simply use CPD four. Both of these options can't create pictures. Let's just test this prompt, a bowl of soap that is also a portal to another dimension, digital art. If we work in the normal CPT four interface, we need to tell CCPD what he has to do. So for example, make a picture about. And then we can simply send this prompt out and CPT will understand that he needs to make a picture about this prompt. This is really key if you work in CPT four because this model is multimodel as you already know. So if you have the po, I would recommend you to work here because here you can be the most precise with this model. And here we also get really, really good pictures because the prompt engineering will be somewhat automatically. Because if you press on this picture, or you can see the prompt engineering was automatically, so we get automatically the right prompt for the optimal output. Now, if you want to make your pictures on your own, so no prompt engineering assistant of Cech BD four, you can totally do this. You simply press on explore, Then you go into this dally and we can run the same prompt and we will get similar outputs but a little bit different. Because right here, Cech PD doesn't help you to create these pictures. This prompts will simply run, and most of the time we get two or three outputs out of one single prompt. This is also really, really cool. If you have something specific in mind and you don't want the Cech BD messes with your prompts. So you can see now we have also these pictures, of course, you can also open this up and see them a bit larger, and the outputs are also really, really good. Besides that, we have two or three other options to use Dali, and they are both completely for free. So the first option you also already know. You can simply go on the copilot of Microsoft, and you can simply tell him. Make a picture so exactly like in GPD four. Make a picture of this bowl right here and you can send it, and then also the Dali model will make a picture. This is also cool because this is completely for free. If you don't have a No, you can and probably should use the copilot to generate your pictures. And here we have it. So four pictures created by Dali three in the Microsoft copilot interface. Besides that, you can use Dali also in image creator from Bing. So this is microsoft B, simply type in Microsoft being image creation, and you can create up to 96 pictures every single day. And if we include the prompt and simply press generate, we will also create these pictures in this interface, and also this is completely for free. And there we have it, so four pictures in microsoft Bing. That are the easiest ways how you can use Dali three, and some of them like two are totally for free. And now I want to show you some other ways because you have so many ways how you can use Dali, you wouldn't even believe this. You need to choose the right option for you. You can use the DI two experiment from PM I also totally for free on the interface. And everything you have to do is to simply log in here, and then you are at labs do open in a polar bear playing se. Create a big robot also right here, but this is the old Dali model at least right now. I hope that they upgrade this because the generations here are not that good, but you can use this for free. So this is the old version. Just keep that in mind. And of course, if you are a developer, you can use Dali three also over the API. You have so many options how you can use Dali three, and also Dali two, this is just enormous. Over the API, you can use Dali three and also Dali two, so the old model. This is enormous, and you just need to understand what you want to create and how much do you want to pay. Besides that, you have also the tools like PO that bring the stuff together. Like I said, I don't like PO. In this video, you have learned that you can use DL over a lot of different ways. If you have the GPD plus bo, I would recommend you to use DALI three in the GPD plus Abo because the prompt engineering is better and the outputs are the best. So just go into C GPD four, the multi model version, and here you can and probably should create your pictures. This is the absolute best way, at least in my mind. You can also press on the Dali bot from PMI if you have the plaza Bok I told you. If you want to make this totally for free, I would recommend you to go into the copilot or into being image creation. The labs don't work perfect because they work with the old models, and if you are a developer, you can also use the API. So choose your right weapon because you can and will create absolutely stunning pictures with Dali two and Dali three. I am 100% ch with that. And of course, we will start into the CCPD interface, and everything that I do in the CPT interface, you can also do in the copilot for free. So if you don't have a subscription, just take these lessons and make them completely for free, into the other interface, into the copilot, or into the Bing image creation, and you are ready to rock with Dali three. 8. Important Prompt Engineering Basics for all Diffusion models Dall E as example: Like I said, Dali is the diffusion model that is easiest to use. And because of this, we start, of course, with Dali. We just need to think about two things, the right prompt and the right aspect ratio, and it depends a bit where you work. If you work in the CCP interface, you can give a specific aspect ratio. If you don't work in the CCP interface, this doesn't work. So you can use different aspect ratios in the CCPD interface, and of course, also in the co pilot because this is basically the same stuff. But if you work into the Bing image creation, you can specify your aspect ratio. So keep that in mind and use your player carefully. I would recommend you to use the CCPD plus or the microsoft co pilot if you want to have a free alternative. But I wanted to bring up every single solution. So in this video, we will make a Dev dip into the prompt engineering because this is the most important stuff if you want to create good pictures. And ironically enough, we go onto the Midourney web page because they describe how you should create prompts for Midourney and also for Dali absolutely phenomenal. So here we have the prompting nodes. They tell us everything. What should the prompting length be? So prompts can be very simple. Even a single word or even an emoji will create pictures. Very short prompts will rely heavily on Mourns default style. And in our case, of course, also of Dali. Because if you don't say stuff, Dali will simply make the stuff up for himself. So a more descriptive prompt is better for a unique look. However, super long prompts aren't always better. Concentrate on the main concepts you want to create. This is key. So you don't want to have a lot of words with no meaning. This is important. So make the prompts as precise and short as possible. This is what they want to say you about the prompting length. This is for every single division model. It doesn't matter in what model you work. Short and precise, but also specific. Grammar. This is something I really like because my grammar and spelling is absolutely miserable. Especially in English. The Michnybt does not understand grammar sentence structure or words like humans. So remember, also the difusion models see word tokens, so they don't really care about how you write a specific word, and they also don't care if a word is not perfect. But word choice also matters. More specific synonyms work better in many circumstances. Instead of big, dry gigantic, enormous or immense. Remove words when possible, so cut the fluff. Fewer words mean each word has more powerful influence. Use commas, prakts and yp pens to help organize your thoughts. But know the Michurney bot will not reliably interpret them. The Micherny bot does not consider hpitalization. Focus on what you want. This is the most important stuff. It's better to describe what you want instead of what you don't want. If you ask for a party with no cake, your image will probably include a cake. If you want to ensure an object is not in the final image. This is for mid journey only, so we skip this. So you simply need to describe exactly what you want to see and don't include the words that you don't want to see in your prompts. This is also true for Dali and every single diffusion model out there. At least for right now, because, of course, we work later in the course with negative prompting, but more on that, of course, later. In Dale, it's not relevant. Think about what details matter. This is the most important stuff for every prompt engineering in every single large language model. Anything left unsaid may surprise you. B as specific or walk as you want. But anything you leave out will be randomized. Being walk is a great way to get varirity, but you may not get the specific details you want. So try to be clear about any context or details that are important to you. Think about the things right here. This is the most important stuff. So you need to include the right subject, the right medium, the environment, the lightning, the color, the mood, and decomposition. Let's make this slow. For example, A subject is, of course, a person, an animal or something else, a location, maybe. The medium is a photo, an illustration or something else. The environment can be underwater in the city in New York, wherever you like. The lightning can be soft, can be neon, it can include studio lights. It can also be the golden hour as explicitly like the golden hour, and the colors, they can be vibrant, they can be, of course, colorful, black and white pastel and much, much more. And you should also probably include the mood if you want to make a picture of persons and animals, for example. Should the person be seeded, calm, energetic, or maybe angry or something else, and of course, include also the composition. For example, should it be a portray, a headshot, a close up, a bird's eye view, a full body view, whatever you like. To make this really, really simple, we make a small example right now for day. So remember, we make our example here. Grammar and spelling is not that important. We should use good words or words that are precise. Of course, we need to include all of these. So we start with the medium, for example. Our illustration of, for example, a CT. We have the medium and the subject. What is the environment? In the garden. We can make commas and so on. Of course, we can also describe how the cat exactly is. Should the cat be cute, for example, an illustration of a cute cat in the garden. What should be the lightning? In the golden hour, the colors vibrant colors. The mood, the cat should be angry. So a cat that is cute but angry. And decomposition full body view. This is a really, really specific prompt, and we can simply copy this prompt, and then we go into Cech PT or B or wherever you want to create your pictures and we will make them. So in this specific example, I go into the Cech Pit interface, so in the normal interface, I use GPD four, and I simply include this prompt and simply say C GPD at the beginning, make a pick off. And then we simply sent this out and we will get our nice little picture. And there we have it, and everything is included. So if you see we have an illustration of the cat, the cat is, of course, cute. She is in the garden. We have the golden hour. The golden hour is, by the way, the last hour before the sun goes under or before the sun rises. This is, by the way, the hour that most photographs love. Some photographs even refuse to take pictures that are not in this golden hour. Yes, this is serious. The colors are, of course, vibrant, so they are really, really shiny and strong. The cat is also angry, and it is a full body view. So we leave no error here. We have everything included. If we simply copy this prompt, we can, of course, also do this in Bing. But remember, in Bing, you can't be that specific with these pictures because after the picture, we can say, for example, make another resolution out of this. Don't worry, we will go over this. But just for reference, if we simply throw this into the Bing image creation tool, we will make a picture that looks nearly the same, and we will get four pictures. And there we have it. So we are really, really specific. We have the illustration of our cute cat in the garden. We are in the golden hour. We have our vibrant colors, and the cat is also angry and it is a full body view. If we delete something out of this prompt, these pictures will vary a lot. If we simply delete the single word cute, I think this picture is completely different. And there you have it. So the cat is not really that cute anymore. It's an illustration. Everything else is right. So the cat is not cute anymore, and the pictures are completely different. And if we delete also the illustration, So we simply say a cat in the garden in the golden hour vibrant colors, the cat should be angry, full body view. You will see that some of these pictures will be, for example, a photo or something else. And there you have it. These are basically all photos right now, and I don't think this cat is really that angry. But just let me know what you think of this cat. So in this video, you have learned the prompt engineering basically of every LLMs. You need to make really, really good and precise prompts because if you don't do that, you will get random results. The prompting length should be precise but short. Grammar is not that important. You need to focus on what matter and don't include the stuff that you don't want to see, and you need to think about what details matter. Think always about the subject, the medium, the environment, the lightning, the color, the mood, and the compositions, and you are ready to rock basically every single division model because this never changes. So this prompt engineering lecture is for every single division model. Please, please remember this because, of course, later, we will include different prompting engineering concepts, but this is the basic. This basics work in every single diffusion model. So please remember this lecture. 9. Midjourney Examples: This video is just a little quick tip. If you ever struggle with finding some inspiration for picture generation, and it doesn't matter in what division model, you should simply go to Murne. And yes, you can also make similar pictures in Dali. Maybe it doesn't work that good, but of course we will also take a really deep dive into Mourny later. So Murne is an enormous webpage. If you simply go to mid churney.com and click on Explore, you can see what people make. And these pictures, I think, at least are enormous. In the Explore deep, there are the best pictures, or, by the way, the pictures that people simply like the most. They don't have to be the best quality, but these are simply pictures that are enormous popular and maybe for a reason. You can see the quality is enormous. This looks for example like your real girl. And also, this looks like a photograph from Aprophy. And you will find a lot of really, really good and cool inspirations. So these pictures at least in my mind, are really, really creative, are really stunning. And if you don't know what you should make, you should simply go into this gallery, and you will find enormous cool pictures. This right here is, for example, like really, really scary, and you find really enormous pictures. The choker also looks cool, and also this astronaut that eats pizza is cool. So this is just a quick one. If you don't know what you should make, just take a look at the mid journey pictures, and you will get a lot of inspiration. 10. Dall-E in ChatGPT: You already learned how you should create your prompts if you want to have a specific output. But if you work in CCPD four, this is sometimes not really needed because, of course, CCPD is the large language model, and Chachi PD will get the context. I think this small video of PMI shows you this perfectly, and then we will also go into Chachi PD because Dali can also do something really, really cool, Dali can also create text inside of your pictures, and you can also download your pictures after you have created them. So, of course, this is simply the article from Dali. You can simply see this article for yourself. You can see all the reserves. You can see the API, the Cech PD stuff, the safety part, and of course, also the information about the company. But what I want to show you right now is a nice little video because sometimes the videos explain it best. So I hope you understand the power of Dali. Di in combination with Chachi Pit understands whole stories. So you can make pictures and then include text. You can write whole books. You can make illustrations just like pm told you in this video. CCPD will understand the context, and then Dali will step in and generate your pictures. Your pictures will be nice, and then you can simply talk to Chachi PT how you want to iterate this if you want to write, for example, a whole. Storyline and include different pictures and much, much more. So the possibilities are endless, and now I will show you that Dali can also generate really, really cool text inside of pictures, and Dali also understands, for example, what's funny or what AM is. This is also relatively new for diffusion models. So right now they understand why am is funny and what am makes am basically. Make a picture of a dog that goes on a walk. And here we have this picture, and let's just assume that I really like this picture. I can download it if I press on this button. So this will get immediately saved on my computer, as you see right here. This is now into my downloads. And of course, I can also rate this. So don't forget reinforcement learning from human feedback. We can like or dislike the picture and simply tell Op My if the output is good or not. I like this output, so I rate it with a thumbs up. And now I will show you that Dali and Chachi Pit also understand the concept and the feelings and all this stuff about these pictures. If we say, for example, make me out of the Chachi Pit will automatically come up with something funny about this picture. So I can also say what Chachi Pit should make funny, but now I want to let Chachi Pit guess so that you can see that this is all possible. So Chachi Pit automatically understands what funny is, and he will also deliver and make a picture that is funny. And he will also understand Damm is fun. Chachi Pit also needs to understand this. And this is perfect because you can also see that words are also possible in Dali. When you hear the word walk and immediately turn into and enter doggo. So you see Chachi Pit automatically understands what Damm. He understands Damm is fan he automatically included also the text. Yes, the text is not always perfect. So here, for example, there is a little bit of a mess up. But generally speaking, Chachi PD and Dali, they do a great job to generate also text, and you can see that Chachi PD understands also what's funny and what not. And you have already seen that we can also, of course, make whole stories. So you can simply tell Chachi PD Hey, make your dog, and then you can tell basically your whole story and simply make it with pictures with text and much, much more. In that way, you can basically also write an entire book. This is really enormous in Dali, and this is also special in Dali because it's combined with a division model. Let's just make one simple example so that you can see the power of these. So you can basically start with something like this. I want to illustrate a book. You help me make the content and the pictures. Give me a picture of a baby fox and a baby rabbit that got friends. Write some lines of text and think how we can go on with the story. So basically, Chachi Pit does everything for us. He makes us these really, really cute pictures. We can make really maybe a children's book out of these. And then we get also some text. I think the text is a bit too harsh or not really for kids. You can also make a story for a bit older people and so on. So here is the story line though in the heart of the emerald meadow and so on. So we can simply also tell Chachi Pit to make this story for like six years old or eight years old, and Chachi Pit will immediately adapt. And then Chachi Pit also tells us what we could do next. As for continuing the story, we can explore how their friendship grows despite their differences. Perhaps they could go on a small adventure helping each other and so on. And now you go simply on. So make this fox and the rabbit, for example, In a house and they got friends because they met a cat and so on, you can and should be simply creative because you can really write entire books. And of course, you can also include text in your pictures. If we go into a new tab, we can do something like this. Make a picture of a dog that says, Go eat that yourself. It should be a funny meme. And of course, grammar and spelling all over the place. This is a really nice spelling for me. And now we have it, so go eat that yourself. I think this picture is also hilarious. This looks really, really nice. So you can see you can also make memes and include text. You can also make something like a picture of a bar where stands open right now. Or something like make a picture of a bar with the words open until you go, so you can really be creative and make whatever you like. And there we have the picture of the bar, and I think also this picture is really, really cool. So open until you go. I think if I would open a bar, I would use a neon sign just like this. Open until you go. That's the right sentences of a bar. So in this video, you have basically learned a lot. You have seen the C CPD in combination with Dali understands whole story lines. They will get what's funny they can make means. They can include text into pictures, and they can help you write stories and much, much more. Also, the small pi video showed you everything. As soon as you have created a picture, you can also download it and rate it for the reinforcement learning. And you have also seen that it's totally possible to write an entire book or maybe make a cool picture or a cool meme with included text. And I think you should immediately go and try all of this out because D is a nice division model. So stay tuned for the next lectures. 11. Different Aspect Ratios in Dall E for Youtube, Instagram or TikTok, X and more: Now you understand the basics of this prompt engineering. So how we should structure a prompt for diffusion model. And you have also seen that we can use Dali in a lot of different environments. Like I said, GPD four is the best option at least right now because we can also include different stuff right here. We can include, for example, different aspect ratios. This is enormous cool because we can simply tell GBD what aspect ratio we want to have our picture in. I found a nice article on this web page, and the web page simply tells us what the normal aspect ratios are. So we have one by one. This is simply a square imatch. Then we have nine by 16. This is vertical, so this is really mobile friendly. For example, for TikTok or Snapchat, and if we switch it, we have the white screen. We have 16 by nine. If I make me a bit bigger, you can see this is 16 by nine. This simply means if we measure these in inches or centimeter or whatever you like. This right here over me is 16 " And this right here is 9 ". And if we flip it, we simply reverse it. So we would have 9 " here and 16 " there. This is basically all that matters. It's simply the ratio between length and height. 16 by nine is also really, really good for YouTube, then we have four by three. This is the standard full screen. So the old TV screens were something like that. And we have also the cinematic white screen. This is 21 by nine, and there are also resolutions. So we have a lot of different resolutions, but to be honest, we can't change the resolution in Dali, so we don't worry about the resolutions. If we scroll down a bit, we have the best aspect ratios for social media on this web page. So you can simply look for yourself, and it depends what you want to make. If you want to make, for example, a Facebook post, You can use one by one or so 16 by nine. If you want to make Facebook stories, nine 16 is maybe better. The Instagram feed, you can use one by one or four five. IGTV or stories should be 16 by nine. TikTok short form nine P 16 or one P one. So you can simply look for yourself, what's the most important stuff for you. If you want to make something from Twitter 16 by nine or one by one, inked in one by one or 16 by nine, and YouTube, nearly all of it is 16 by nine, also if you want to make a thumb nail or whatever it is. Of course, if you want to make shorts, we should choose nine by 16. But right now we don't make any shorts and we also don't make videos because we stick with picture at least right now. Later in the course, we will also see how we can make videos, for example, for social media. But like I said right now, we stick with our pictures. And now I show you how we can use this in day to make different aspect ratios out of these. We use GPD four and we go into a new tab. And now I will make a picture, and then we will make this picture in different aspect ratios. I simply tell make a picture of a mystical illustration of a red dragon. So you see what we do. We include everything in this prompt. Subject, medium environment, lightning color mood and compositions. We have a dragon. This is the subject. The colors are red. We make an illustration. The dragon is also mystical. The environment is in the air. The lights are shiny. We have our full body view, and the dragon is calm. And then I simply also tell CCPD aspect ratio, 16 by nine. So you can use this and maybe make a YouTube thumbnail out of these. And there we have it. So this is 16 by nine, perfect, for example, for a YouTube thumbnail or also if you have a screen like this screen right here, you can use this on your computer, and this is simply me, but just scaled up. So I am right here. In this interface also 16 by nine, if you scale this up, you see that the ratio is basically the same if we scale this up. And of course we can and should change this if we want. If you don't want to make a YouTube thumbnail, but maybe something for an Instagram story, you can simply change this and we flip it. So we make nine by 16, and then we get our picture for an Instagram story, for example. So this would be perfect for an Instagram story. And if you want to post this in the Instagram feed, you can also use four by five, for example. This is also really, really nice if you want to make a post into the Instagram feed. And this is basically four five, but I must admit I am not entirely sure if this aspect ratio is completely true. So 16 nine and nine 16 is perfect in C CPD and in Dali. I must admit that the aspectrato doesn't work all the time. If we delete this one less time, and I make here without any aspect ratio, we will get a square picture and square pictures you can louse nearly all the time. And there, we have our little picture in one by one. I have to admit that in Dali the aspect ratio is not that accurate and not that perfect. In mid journey and also in stable fusion and the Dobe firefly, we can adjust the aspect ratio really, really precise and perfect. And we have, of course, also 21 by nine. I don't want to show you also this right now because I think you get what I mean. So in this video, you have learned what the aspect ratio is. The aspect ratio is simply who has fought of this the ratio between length and height. And you can use different aspect ratio, for example, for different platforms. If you want to make something for Facebook one by one 16 by nine works good if you want to make a post. Stories are in nine 16 perfect. The insta feed is in one by one or four five perfect. IGTV is 16 by nine. If you want to make something for DTok nine by 16 and one by one is perfect for Twitter, 16 by nine or one by one, Linktin one by one, or 16 by nine Linktin stories, nine 16, 16 by nine for YouTube. And if you want to make shorts, nine 16. So if I had to make your summary, what aspect ratio you should use on what platform? If you make something like short content, so stories, and so on, use nine by 16 like because everybody looks on this phone right here. If you want to make longer form contact thumbnails and much, much more, do it in this way. So use 16 by nine. And that's basically it about the aspect ratio. Like I said, we can use this in Dali, but only right now, at least if we work in CCPD and we simply can tell CCPD what aspect ratio Dali should generate. This is not always perfect because four by five seems to not work all the time, but 16 by 9916 and one by one is perfect, and you can do nearly everything with this three aspect ratios. So have fun trying this out if you work in CBD plus. And by the way, in the other diffusion models, the aspect ratio are really really dialed down and precise. In Mid journey, for example, you can also use 128 by 96. Md Journey will simply count the pixels. But like I said, we start with the simplest difusion model, and that's why we start with D and of course, also know CCPD, and for that reason, we start with DL just like I told you. 12. Taking reference pictures to make similar pics in Dall E: In this video, I want to show you how we can use reference pictures to make similar pictures in CCPD with Di. Who doesn't know this problem? You see a picture somewhere, but of course, they are copyright protected. You can use them or whatever. Or you simply have a picture and you want to make more variations out of this picture. In this simple example, I will show you how we can take this picture. We can feed it into CPD and we can create similar stuff. For example, for your web page or whatever. Now, let's assume you want to make a similar picture than these. Maybe you have a web page and then you are a chiropracticer, or however they are called in English. You simply want to tell people that they should be careful with their knees. So we simply upload this picture, and then we can make similar ones. We press on this button right here. Then I upload this picture, and we can use two different options. The first option we will use Cech B division to describe this picture and then tell Dali to make a similar one or we go instantly with the way to create this picture. Let's just try the instant way first and see what we get. So I start with make a similar picture, and you see Dali will get used instantly, and there we have it. You see the picture is nearly the same. We have a really good consistency. This is the easiest way how you can create similar pictures. Yes, of course, if we go into a new chat, I will also show you the second way. Of course, we'll load the picture up and I simply ask CCPD to describe it. Okay. So describe this picture. Now CCPD has a lot of context, and then we can simply tell CCPD to make something similar. So make a similar picture. So you see first, we used vision and the code interpreter or advanced data analysis, and now we use d to create a similar picture. And there we have it, and I think this is nearly also the same picture from the previous chat. So they are really, really similar, and I think they are perfect. They are just like the picture that we want to make. And of course, we can also go on and we can simply tell ChatBD that he should fix some things. So we can maybe say that he should mark the back and not the knee. So make a similar picture, but mark the back with red color. The knee is healthy. And there we have it, and this was not entirely perfect, so of course you need to play a little bit. Yes, the back is now marked with the red color, but also the knee is marked like a little bit. We can go on and chat with GPD what we want to change, and then we get a really, really good output. Of course, if we want to make just small tiny changes, we can do this via the N ID, but more on the chen ID in the next video because I think the chen ID is genius. In this video, I wanted to show you how you can use a reference picture to make similar pictures. Most of the time you see a picture somewhere and you think you need a similar picture. You can simply use CCPD to analyze the picture and then create something similar with Dali. With most pictures, this works really, really well, and then you can simply change different stuff if you simply tell CCPD what he should change. And of course, you can make this even more precise with the GND. And about the Gen ID, we will talk in the next video because with the Gen ID, we can also create really, really consistent characters. So stay tuned for the NIE because this is key in the A. 13. Use the Gen ID to create Consistent Character: In Dali, we have something that is called the Gen ID. The GNI is basically the same stuff as the seat, but a little bit different. And I'm sure you remember what the seat is. The seat is the first starting point of the division model, and you remember that we can make consistent characters with seats. And the same thing is true for these en IDs. If we simply tell CCPD or Dali to make something with a specific GeniD we can and will make really really consistent pictures and can change just minor things. So let's just take a look how we can do this. First of all, of course, we need to create a picture. And we can and probably should make this relatively easy. So we simply make something to start, and then we change the starting thing and make it really, really similar. This is really key if you want to have a consistent character, if you want to write a book, for example, this is enormous, powerful. This right here is my first prompt. And remember, we work in Cache PD. And if we don't work in Cache PD, you can delete this, so you can delete, so you can delete, give me. If you make this in being image creation or in another tool, you can simply start with this Illustration of a ustical graceful deer with alers that glow like the aurora borealis. I think this will make a really awesome, cool picture, and you see that I take most of the time illustrations for D. Because yes, Daly can make photo, Daly can make really realistic pictures, but Dali is better with illustrations. In the next video, I will show you also how we can make really, really realistic pictures because for that, we need to make a really, really cool prompt engineering trick. So right now, I simply sent this out, and I am sure we will get a really cool picture. And there we have it. I think this is a really, really cool picture. And what we do next. Let's just assume you want to have this picture maybe in the summer or in the deep winter with a lot of snow or whatever you like. I would do something like this. Just remember, if you simply tell CPT right now, make this picture, for example, in a deep winter, yes, we will get a picture p. The picture will not be that similar. Let's just test this out, and then I will give you the solution. So make this picture but in winter with snow. And here we have the picture, and I must admit, yes, the pictures are relatively similar, but they are, of course, not perfect. And now I will show you how we can make pictures that are even more similar. If we simply copy this prompt and go into a new chat, I will show you exactly everything that you need to do. So we make a picture again and we will get a slightly different picture, of course. Here we have our picture, and now I tell CCPD something like this. Give me the chen ID for the picture. So the chin ID is this right here, and now I simply copy the chen ID and I tell Chachi PT, give me now the chen ID but in winter. And now we will get a picture that is more similar to the previous one. So the picture will be relatively similar than this one but of course in winter. As you see right now, we have some flowers and so on. The picture is really, really cool, and now we will get a picture that is really similar but of course in winter. And perfect. So yes, this is really nice. There's some snow on this whole stuff, and this looks incredible, at least in my mind. Look at what we do right now. I simply copy these and now I tell CCPD something really cool. So give me exactly this ID, but not in winter but as a baby. Then we will get a picture that is really, really awesome because we will get a baby picture of this deer. This is perfect. Of course, this is not in winter because I simply used the first ID. So this is the first picture, and this is the picture when this thing was a baby. And we can do all kinds of stuff. I can also do something like not in winter, not as a baby, but as Grandpa, for example, and we make the grandpa of this little dear. And here is the grandpa of this deer and all thanks to the chana de. The Cena is awesome. And yes, this example was maybe not perfect because a deer is always a dear. But if we take a character, maybe you are really writing a book, then the Cena is even more powerful. And now I will show you how and why. Make a cartoon style animated illustration of a young girl with pink hair and big green eyes that is wearing a purple dress. And there we have it. The picture looks really nice, I think. And this is now key because here we can change minor things, something like the clothing, and we will always get a relatively consistent character. Give me the chain ID of the picture. Here we have the chen ID. Then of course, I tell Chachi PD the same stuff. Give me then the chen ID but with a red dress, and then we should get theoretically a relatively similar picture, but the girl should be in a red dress. And you can see, these are basically really, really similar pictures. So you see the consistency is enormous. And then I can do, of course, the same stuff. So I simply copy these and then not in a red dress, but something like maybe a green dress or whatever you want. But with a blue dress. So you can simply change outfits or maybe change hair color or change hair length or change different stuff by using this gende but the general character is really, really similar. You see this is really key. We get really similar, but with minor differences. If you want to write a book, for example, this would really work at least in my mind, because the pictures are really, really similar. And here we have our picture with the blue dress, and the character is always relatively similar. Yes, the hair length is not perfect, but maybe you should simply run the prompt a few times, and then you can really get good good outputs. I think the outputs are enormous, and if you don't use this GID, you will get a little bit more random resorts. Just make one and now we make one last split test. We simply start with this prompt, and then we ask CCPD to make the same picture. But for example, with our blue dress, and then we see what we get if we don't use the ID because we get a lot more consistency if we use this ID. So here we have our picture, and now I simply ask, give me the picture, but with a blue dress. And then you will see we have, yes, we have a similar picture, but the picture will be more different than we use a Gen ID. And there we have it. So you see the complete face is different. Also, the mouth seems to be different. The colors of the eye browns are different. These are eye purpose or however you want to call them. So this is basically not the same picture. The Cena is simply just like the sat in other difusion models. So think of the Cena just like the seat but specifically for Dali. So in this video, you have learned that the chana D is simply the same thing as the sat in stable difusion or a miourney. The Cena D is the first starting noise of the diffusion model of Dali in Cache PD. And if you combine the gen ID with specific words, you can simply change minor things in pictures. This is key if you need a consistent character. You can make peoples or persons environments, animals, and much, much more. And if you ask for the cheniD you can make minor changes. This is key if you want to write something where you need a consistent character. So this was the enaD and I hope it is useful for you. 14. Make Perfect Prompts for realistic Pictures with Cameras and Camera Lenses ChatGPT can help: In this video, I want to give you whole tricks to make prompt engineering a lot better. After this lecture, you will know how you can make pictures that are a lot more realistic and this works in every single difusion model. As you know, you should always describe these things. So subject medium environment, lightning, color, mood, and composition. This is not new for you. But what it's new for you is that you probably should also include cameras, camera lenses and much, much more if you want to make a really specific eye catching picture. This gets enormous better if you do that. If you make, for example, a stock photography or a picture of a dish or something like that. The output will be relatively good, but it will not be optimal, because you will get optimal outputs if you use cameras and camera lenses. And if you think a bit, you also know why? Because all these difusion models are trained on a large amount of text and a large amount of pictures. So they feed pictures into the difusion models and they describe it as precisely as possible. If they feed photos and the photo was taken with a specific camera and camera lenses, then the description will also be that way. And if we want to make a picture that looks like a picture from a photographer, we need to include cameras and camera lenses. And don't worry if you don't know nothing about cameras because I also don't know nothing about cameras. But do you know who knows something about cameras? Chachi BT knows everything about cameras. Here I'll show you a quick example what I did in CCPD to make something really, really good for my prompt. You are a prophy photographer. So first, I give CCPD a role, and that's basically all we need to do, and then we give our instructions. So this is the combination of the role and instruction prompting. I think you already know this from the LLM section. You try to make a perfect close up picture of a delicious dish. What camera with what objectives and lenses do you use. And then CCPD tells us basically as a professional photographer aiming to capture a perfect close up and so on. He tells us a bunch of stuff about cameras. He tells us a bunch about lenses and accessories and much, much more. But of course, this is too long for our prompt, because you already know that we should be short and precise, if possible. And then I simply ask CPD describe the camera with objectives and lenses in as few words as possible. Take the best one. And then we get it. On EOS five D Mark five with F 100 millimeter F divided by 2.8 lens macros US lens. I hope this is somewhat correct. But I think this will be a nice lens to make our nice little picture. And now we will do this. First, we create a normal picture of a stake, for example, and then we include this and we see what's the difference. First, not good prompt engineering. And remember, I only used the word stake. This is not a perfect prompt. I want to make this simple that you see that the camera makes a lot of difference. Make a picture of a stake, and that's basically what we get. Yes, the picture is relatively good, but this is not optimal. And now compare it to this. Make a picture of a stake. The photo is taken with con, EOS, and so on, and you see the difference. So the difference, at least in my mind, is enormous. And now I try the same thing, of course, also into this being image creation tool. This is also perfectly fine. So first, I make simply the prompt take, and then I include the camera lenses. So a stake, the photo was taken with con and so on, and then we simply press create. And I think you get what's the difference. The steak looks delicious. The steak looks a lot better. The background has an enormous quality. This looks really like a photo with such a lens, at least in my mind. And of course, that's not all. We can simply ask CCPD if we want to make a picture, for example, for a close up phase. And now we basically have the same camera but with different lenses. So I think this should also work. Let's just test this out. I copy this, and then we go back into CCPD, and now we ask for a close up of a woman. So first, close up of a beautiful woman studioize she is calm. And now I simply use the same prompt but, of course, with this camera. So we copy both of these, and I think the output will be a lot better. So close up of a beautiful woman, Studio lights. She is calm. Photo is taken with this cane on and so on, and now we send this up, and then we see which picture is the better one. And there is our new picture, and I think the new picture is a lot better. So in this video, you have learned that you should probably include cameras and camera lenses in your prompts. And you can use CCPD to write your prompts for the cameras, and of course, also the camera lenses because I myself, I am not a photographer, but we can use CCPD to write this prompts to make all of this a lot better. And in the next lecture, I will show you the super trick and you can simply copy and paste the super trick into the custom instructions, and then you will make enormous good pictures, so stay tuned. 15. Dall E HACK, Use Custom Instructions for perfect Pictures in no Time: In this video, I will show you a super trick. We can use C CPD and the custom instructions to make the outputs of Dali even better and even a lot faster, and we will even get more inspiration for our next pictures. This lecture is awesome because we combine division models and large language models, and by the way, this will not be the last lecture where we do that. So first of all, you already know the CCPD kind of helps you with creating your prompts if you make just a normal prompt inside of Dali. Let's just make one simple example right here. A picture of a dessert. Of course, CCPD will create a picture of a dessert, and the dessert will also look relatively nice because Cache PD will automatically generate a prompt that will make better pictures. If we copy this prompt and simply go into the Bing image creation, we can also make a picture, but the picture will not look that good. So we delete here our old picture, and I simply generate a new one, and then we see if Bing or Chachi PT makes the better picture. Normally Chachi PT should make the better picture. So you see, yes, this is a nice little dessert, and I think the picture is also relatively good. But Chachi PT doesn't use this prompt to generate this picture. The prompt from Chachi PT looks completely different and you can see the prompt of Cech PT if you simply press on this picture. You can simply press here. And here, you basically see the full prompt, a picture of a delicious and visually appearing dessert. Imagine a slice of rich and so on. So you see we get a nice little picture because GPT uses a better prompt. And you can see the picture in the Microsoft Bing image creation tool is kind of funny. Yes, sometimes also this tool makes really, really good pictures, but sometimes to be honest, not really. And you need to simply guess if you don't make specific prompts. But all of this has an end right now. Because now I will show you how we can prime C CPD to make perfect prompts for Dali, and all of this will be instantly. You can simply copy my custom instructions, and then you are ready to rock. And this right here are my custom instructions. Right now, I simply copy and paste them, and then we see what we did right here. This is the first thing that I like to copy, and then we simply press on the custom instructions. You already know this. Of course, we need to enable them first, then delete the old custom instructions. So I have deleted all of this. Now I simply copy the instructions, and I paste them right here in. And if there is a little bit too much information, I simply delete the last few words. And then I do the same stuff on how would you like CCPD to respond. So first, this was, of course, what would you like CC PT to know about you. So this is basically this right here, and then we scroll down. And this right here, how would you like C CPT to respond. We simply need to copy also these, and then we paste it into CPT, into the custom instructions, we delete the old one, and then we are completely ready to rock. I click safe, and then we are set for perfect dessert. So I copy this prompt. Now I go into a new chat. Then we make one last Check if the custom instructions are included, they are and see what picture we get right now because I would assume this picture will be awesome. And there we have our picture, and I think this picture got a lot better, and why got this picture better? Because we used a better prompt. If we simply press on this picture, you will see what we have. We have in the prompt included a lot of cool stuff and also camera lenses, camera settings, and much, much more and really, really good lightning. And this is basically the key. And if you see right here, we get also some information, what we can generate next, a close up photo of a vibrant fruit tartar, an elegant plate chocolate suffle rustic apple pie on wooden table, a colorful and so on. I simply want to make this elegant chocolate suffle maybe after this also the vibrant fruit salad tartar or whatever you like. You get simply some inspiration, what you can or should create next, and all of this because of our prompting. Awesome. The chocolate souffle looks at least in my mind even better because this is like in a restaurant, this looks really, really awesome. Thanks to the camera lenses, and all of this, this works really, really nice. So we can, of course, rate this picture just like I told you, with a thumbs up because I think this is really nice. And I will also download this picture right here because I think this is a cool picture. So we download it. And then we get for other prompts, but I think right now you get that we will make really, really good pictures with this custom instructions. Let's just take a look at the custom instructions why they are so powerful. First, I give prompt generation guidelines, so create prompts that paint a clear picture of image generation. Use precise visual descriptions, rather than metaphorical concepts. Try to keep prompts short yet precise and eye inspiring. Then the structure. So a medium subject subjects characteristic, relations to background, the background, the details of the background, interactions with color and lightning and then taken with or drawn with a specific trail list of style. And that's why we include cameras and so on. Then we have the medium consider what form of art this image should be simulating. And the object, so we simply describe everything about the stuff that we want in the picture. So what is the main focus, the colors, the pose, the viewing angle, the background, the environment. So is it indoor, outdoor, abstract, and so on, the colors, the lightning, what time of the day intensity, and so on, and then also the style traits. So what are the unique artistic characteristics? What are the influences? The techniques? Is it, for example, digital art or is it a photo? And if it's a photo, we should include photographic camera gear and camera settings. This is really, really enormous, good and powerful because we can simply include one word and get everything that we need. If it should be a painting, he should also describe texture and brush strokes. Is it digital note the software used? This is basically everything that we post in the first chat. And in the second chat, we simply tell CCPD how he should behave. So how would you like CCPD to respond? Generate images based on your detailed prompts using the AL free. I always bring the idea I life with bold and interesting choices for every element of the prompt, always follow the prompt guidelines. Suggest four brand new ideas that I can rip off. So this is the stuff that's also really cool because sometimes we don't have the creativity to make new prompts, but Dali and Cech BD will deliver new stuff. This should be simple concepts and not full prompts. Try to take inspiration from the last suggestions I gave you rather than the full prompt. And that's it. I don't need any further context. The less fluff you include around the generations, the faster I will see the images and be able to intergrate my ideas. And then, of course, also the defaults. So the default means simply what are the default settings? So the aspect ratio is simply every time one by one, unless with CCPD to take another setting. It should be always a photo. Unless with it should be digital art or something like this. And CPT should always make for pictures But right now, this seems not to work because I think a lot of people use D and they don't have enough GPU power. And then, of course, we have unimportant information. Avoid words or concept that go against terms of service. Do not infringe on anyone's copyright. Do not use suggestive or explicit imagery in your prompts. Do not emphasize or imply any elements that would not be considered G rated. And that's basically it. That's the whole secret. With all these custom instructions, you get enormous good pictures, and everything you have to do is to simply write a small little prompt. So something like this creamer. Then we can simply include it and send it out, and CCPD will generate a really, really good and nice picture, and it will always be a photo in one by one, and camera lenses will get included, and the background will be described really, really good. So everything will be completely on point thanks to this custom instructions, and I think this custom instructions are completely gold because you can save a lot of time. Yes, CCPD will also help you without the custom instructions to make relatively nice pictures, but the pictures are not that good than these pictures. Yes, these cram bros maybe not optimal. I think cram bros never much prettier than these, but this right here, this looks really optimal, at least in my mind. So in this video, you have learned that you can hack the system. You can combine LLMs and division models with this custom instructions. If we tell JGBD in the custom instructions how he should behave, he will generate automatically enormous good pictures with Daly and he will include camera lenses and much, much more with really, really short and precise prompts. I think you should totally do this, so just simply copy my stuff and make enormous good pictures. And after every picture, you will get four suggestions what you can do next, and that's probably what you should do next. 16. Conclusions of Dall E in ChatGPT and a bit of Homework: Over the last videos, you have learned a lot. You have seen how you can get access to Dali. You can use CPD. You have two different versions in CCPD, the multimodel version and of course, also the D CPD. You can use Microsoft copilot if you want to make the pictures completely for free, and in the copilot, we work exactly the same as in CCPD. You can also use the Bing image creation to make your pictures, and also this works relatively nice, but B doesn't help you to create your prompts. And we have also Dali two in pi labs. This doesn't work optimal because we use at least right now an older version. Then you have learned everything that is important in prompt engineering. So you know how to craft really, really good prompts in Dali and also in other diffusion models. You have also seen that CCPD understands whole story lines, means and much, much more and can even generate text into pictures. You understand what aspect ratio you should use for what situation and how you can make them in CCPD. We have uploaded reference pictures and make similar stuff. We used the ID to make consistent characters. We included camera lenses and all specific camera settings for perfect pictures. And then you also learned how you can combine LLMs and division models, so GBD and Dali to write perfect prompts all with the custom instructions, and you can simply copy and base my stuff. So you basically understand how you can and probably should use Dali perfect in C GPD four. And I would really recommend you that you should play a little bit. At least try the prompt engineering, and if you like to work for free, try all of this in the Microsoft copilot. So have fun trying this out, and if you have GPD four, just copy and bate my custom instructions and you will make amazing pictures. So have fun exploring all of these, and in the next section, I want to talk about the OPI labs because we can also use DI to edit our pictures, and that is also really, really cool. 17. Adobe Firefly, how to get access and what is the difference to Dall E Stable Diffusion and Midjorney: This section is all about Adobe firefly. And you already know what Adobe firefly is. Adobe firefly is another diffusion model, so we can basically generate pictures. But we can't only generate pictures. We can also edit these pictures, just like in the p Mi labs from D. So you can generate pictures and you can edit them. And especially the editing is on the next level with Adobe firefly. So the in painting and out painting works really, really good. We can also completely remove background, make text effects, and also do generative free color. So adobe firefly is a really, really cool, difusion model, and you can use it or at least test it completely for free. You get free credits every month, at least some free credits. Adobe Firefly is also different than all the other difusion models, and these are here the three key points. The advantage of generating AI art with Adobe firefly. This is, by the way, an article directly from Madobe. So Firefly creates beautiful artistic and photo realistic images based on simple text prompts, type into a browser based interface, perfect to speed up creative workflow. So it works similar than Dali, but the interface is a bit nicer. We can make different aspect ratios and much much more. Once firefly generates, images you can adjust the result by adding more descriptions to the prompt or by using tools to tweak the style, theme, lightning and composition. So Firefly is really, really easy and nice to use, and the coolest thing is this right here. Designed to be safe for commercial use. Firefly is trained on Adobe stock, open licensed content, and public domain content where copyright expired. So this is key right here. Adobe Firefly is easy to use can make good generations. You can test it for free. And the coolest part is you don't have to worry about copyright because Firefly is only trained on pictures where copyright is either expired or of course, on pictures of Adobe stock, where they have the copyrights. So all of this should be completely safe to work. Don't worry about copyrights. And now let's see how you can get access. You go simply on the Firefly webpage. Just typing in the browser, Adobe firefly, and the first link is this, and then you click Get Firefly for free. And this is basically the web page. So you can make text to image. You can make generative feel so the in painting. You can make text effects. You can also do the generative recolor. You can do text to template and also text to vector graphic. So you can do a lot of stuff. You can even use the upscaler and also this is really, really nice. Adobe firefly is an enormous tool, and I think you should totally know this tool. And this is the stuff that will come in soon. So free D to image, sketch to image and also personalized results. They do a lot of stuff that also stable difusion does. In stable diffusion, we can already do free to image and also sketch to image. But I think Adobe fire fly is a really, really nice tool and you can and probably should use this at least right now also totally for free. And what you need to do to make this for free, just login with your Google account, and then you are on this web page and you have every single month 25 free creations. So you can simply play a little bit with this tool. And by the way, if you have an Adobe account where you pay. So if you have the subscription for Adobe, you can use Adobe Firefly completely for free. This is also really, really nice. So if you already have an account in Adobe firefly, you can use it for free. And if you don't have an account and want to upgrade it or use more than 25 generations of firefly, you can also totally do this. Everything you have to do is to click Upgrade here on this button. And then you can simply make your plan. So you can use the premium plan of Adobe firefly, or you can use the premium plan of Adobe Express. I must admit I personally had the premium plan of Adobe Express, but right now it's expired, and I don't pay this right now because I don't use this all the time. So for me right now, it's not really worth to make your subscription to this tool. But if you want to use this more, I think you should totally do this. Just start for free and see where you go. So in this video, you have learned what the Adobe Firefly is. It's a division model, but it's trained on pictures of Adobe stock and where the copyright is expired. This is really, really cool and nice. You can simply go onto their web page and logging possibly via your Google Coto, and you can try all of this totally for free. In the next video, I will show you the interface and how everything works because, of course, you're here to make pictures. 18. Adobe Firefly Platform Overview and first Picture: In this video, I want to show you the standard interface of Adobe firefly. It's really, really easy to use, and then we will create our first picture, and of course, the prompt engineering is really, really similar to the stuff that we have already learned. So the interface of Adobe firefly. Like I said here, we have home. Then if you press on gallery, you can simply see what other people have made and you can get inspiration for what you can make. You can simply see for yourself. These, for example, are the text effects. So you can also make really, really cool texts with different effects. Then you can also make something like animals or cars. And you should simply scroll down a little bit and see if there's something that you really like. And if there is something that you really like, you can simply press on it and recreate similar stuff. For example, I really, really love this picture because I love to eat seeds. And I think also this picture looks really, really realistic. So if you find something that you love, just press on it and you will see what's the prompt. Here you are, for example, immediately into the image generation tool, and you see what the prompt was, and of course, also, what the version was and all of this. Don't worry. We make a deep dive on all of this so that you understand how you can create this pictures. You can see we have right now created four really really good and cool pictures, and all of this by this prompt because we have copied this prompt. If you want to go back, you can simply press back right here, and then you are again in the gallery, and if you press back one more time, you are back here. If you press on favorites Here is simply just the stuff that you have saved. You see right now I have only one picture in my favorites here. If you click on about, you are on the general web page of BFirefly and you can simply see for yourself how they describe everything. And then, of course, you also have the help button. You can press the help button and see the entire FAQ. Here is also stuff about the copyrights and much, much more. You can simply see everything in this FAQ if you like. But I think after the course, you don't have any questions left. If you go back on home, you see also that you can either upgrade or you can press on this button. And here, though, we will give you some examples how you should use their tools. So Firefly image. To model is aaable right now, and you get always some good information. So what's the generative match, for example, how the photo settings work, how you should make your prompts or how Firefly helps. You create better prompts, and also some stuff about negative prompting. If you have more problems, you can press on this and you can report, for example, ideas or a bag or something else. And you can also join their community, and I would recommend this. And this right here is your profile. If you press on it, you see your own name, and you see how many credits you have left for this specific month. And then there are basically the different tools that you can use. For example, text to image, creative fill, and so on, I already showed you this. And if you press, of course, on text to image, you simply have to press generate Then you are again into this community showcase. And if you press on one of these images, you can use the prompt, and if you don't want to use one of the images, you can also describe your prompt right here. If you type in, for example, frog, you can make your first picture if you simply press generate. But I have to admit this will not be a perfect picture. But now we will create this frog, and with one simple click, we did it. And here we have our rock, please don't ask me what the rock does in the first picture. And here we have always the same stuff. So reinforcement learning is for every single LLM and also for every single difusion model. So if you like a picture, you can give thumbs up. If you don't like a picture thumbs down, and if you want to report a picture, you can also do that. And you can edit these pictures. You can send it. You can also save it to favorites. And in the next video, I will show you how we can use all of this on the side bar because there are a few tricks, how we can use adb firefly to generating really, really good and cool pictures. So in this video, you have seen the standard interface of Adobe firefly. The easiest way is to simply go into the gallery showcase and try a prompt that you like. And you have also seen that you can join the community and basically you have seen what every single button does. And now we will discover every single button in the image generation because this works really, really easy. And I think you should know how it works. 19. Generating Pictures in Adobe Firefly, the basics and Prompt Engineering: In this video, I want to talk about the image generation in a firefly because the image generation in Firefly works easy and relatively good. So you already know you can simply search here for different pictures. And let's just assume that you really like these oles because at least I like them. We can simply press on them, and then we copy the prompt. So the prompt is now included into our prompting box. So chicken night o standing next to each other. Each ole has features with a specific colors the rainbow made of yarn. So you see this is the prompt, and if you think about prompt engineering in Adobe Firefly, always think about the stuff that we already learned. So think about what details matter. Include subject, medium environment, lightning, color, mood, and compositions. Always think about this stuff. Even if you copy a prompt and the prompt is relatively good, always think about what you want to see, think about what animal is, what medium and so on. I think you get what I want to tell you. This is about the prompt engineering. But of course, the DV firefly has more to offer. We can and probably should use first of all the right model version. If you click on this, you see what model you use for your generations. We have the first model and the second model. And you always see for what the model is, generate stylized photos and illustration, the FireFly version one and the version two, generate higher quality images of people, improved text alignment, and enjoy better style support. Generally speaking, I would say just work with the version two for everything that you do. And of course, this is really, really cool because the version two is also the default setting. If you have made your picture, let's just assume this was your prompt or you really like this prompt and you like this holes. You can also use different despxratios, and you already learned for what despxratios are. So if you want to make your landscape, this is the thing you want to make. So four to three, then you can use three to four. This is something like for Instagram post square one by one. This works nearly all the time and screen 16 by nine. Right now, these are the aspect ratios that you can use in Adobe firefly. Let's just test 16 by nine. And as soon as we use 16 by nine, we can simply press generate, and then we will get our holes in a different aspect ratio. And there we have it. Now the les are in a different aspect ratio. And you see how fast all of this works. And then the next button is the content type. So either you can use automatic. So this will choose automatically what content type it should be, or you don't use the automatic stuff, and you simply should it be a photo or should it be art, and the visual intensity you can also alter. So if you use digital art, you get something like this. And if you use, for example, this photo, you can also change the style really, really quick, and you can increase or decrease the strength of the different types. Let's just test this one more time. This time with photo and a strong, strong intensity, and we press generate. And you see the picture looks a lot more like a photography. This works really, really fine and really, really nice. And then if we scroll down a bit, you can also see this right here. You can reference images, and you can also upload different images. This stuff, I will show you in the next video because this is a bit special. If you scroll down a bit more, you can also include different effects. You can use up to three different effects. You can include, for example, digital art, photo and painting, and this will always be included right here in the prompt. So right now, we have included photo, digital art, and painting, and of course, none of these makes sense. If we want to include something, we should maybe take photo and maybe hyper realistic, so the painting we don't use the painting, but hyper realistic and digital art, we also don't use digital art. So this would maybe make sense. Photo and typer realistic. And for this generation, I will change the prompt, because, of course, this is not perfect for a photo. So let's just make for example, a stock photography. If I simply type in stock, and then photography, and you will automatically get suggestions from Adobe firefly what you should use. So stock photo of abstract background with purple and green colors. Stock photo of a beautiful woman on top of a mountain, stock photo of a person walking alone in the forest. And we get relatively good prompts. Let's just use this beautiful woman. And of course, you can and probably should also improve this. If you press one more time again, Then you get a lot more stuff to include. Summer or spring tourism concept, for example, you can simply include a lot of cool stuff right here. Every time you press on these, you get new stuff that you can and probably should include. And of course, it is not the goal to make the prompt extremely long and extremely unspecific always think about this stuff. So we have a girl. It is a photo. The environment is in the mountain. Lightning, we need to see if the lightning is included, then color mood and compositions. So maybe we could also include lightning and the composition. And if you don't want to do anything, fancy, just go with our normal prompt. But now we have included also the photo and the hyper realistic. And I think we will get really good outputs just because we have included photo and hyper realistic. You don't have to make gigantic prompts. If you don't include the stuff that is included right here, you will get a bit more random results. And that's also not that tragic. If you don't say, for example, what lightning it should be, the lightning will just be random. And random is not bad. Random is just random. It's just not specific. So if you don't tell the diffusion model that you want black and white, you get it. And if you don't tell the difusion model that you want a headshot or a portray you want, get it. The diffusion model will deside for you. And that is not bad, like I said, because right now, I think we have, yeah, maybe unacceptable photo. So the stuff in the right corner is somehow acceptable, but it is really, really, not perfect. And the first picture right here, this is really awful. And also the face of the second one is not really that good. So stock photographs, you have a little bit to play with it. Sometimes it works really good and sometimes it doesn't work that good. The environment, I have to admit the environment is enormous good. So the first picture, if you look just at the environment, the environment is perfect, but the woman is not perfect right now. So I would say we give here a thumps down because this is just simply not good output, at least in my mind. The face on this picture is also not that great. All the rest is relatively good, but the face is not perfect. Of course, you have to simply play a little bit with all this prompts. You can always delete this stuff and simply make new prompts if you don't like this. But right now, I want to show you that we also have different stuff. So color tone, photo settings, and the advanced settings. All of this, you can simply press on all of this, and then you can include this. So you don't even have to make really, really good prompts. You can simply type it in right here. So illustration of a CT, for example, and then we use simply automatic right here. We scroll down, so the visual intensity, I would say medium, the style strength a bit higher than medium, then we can go down. So what effects do we want to use maybe digital art hyperrealism. We don't use this. We don't need it. And then the color and tone. If we press on the first one, we can see if we want to have black and white, cool tone, gold, muted color, pastel colors, and much, much more. I think we use a warm tone. Then the lightning. The lightning should be somehow dramatic. Dramatic and warm. This is maybe not perfect. So let's just use something other Turia lightning. This is cool. And then the compositions. So what's the composition? It is, white angle. And that's why we don't have to include all of this into our prompt, because we can use colors, lightnings and composition all on this right side. And if we go into mid journey so color, lightning and composition. So these three things, you don't have to take necessarily into account when you write the prompt for ado firefly, because you can simply include it in this simple box. And then you have the advanced settings. This is basically the negative prompting, I will show you also the negative prompting in the next video. And now we simply press generate, we have an illustration of a cat. It will be digital art. It should have a warm tone. It has surreal lightning and it is a wide angle. So we press generate. And here we have the illustration of a cat, and I think this last picture, this is really, really cool. And I will give this a thumbs up. This is an enormous cool picture. And that's basically it with this image generation. You have seen this is really, really easy. In the next video, I will show you the last things because, of course, you should also take an image to start, and you should also use the advanced settings, but this is the standard. And remember, you can always go back. And search for inspiration on this showcase. And if you ever find a stoke photography that is really, really cool, just simply press on it and see exactly how the prompt is written. So this is, for example, a medium shot of a happy child unwrapping a present. And I think this picture got really, really good, and we simply used art, and nothing else is in this prompt. You can see even the fingers are relatively good. They nearly look like a photograph, at least in my mind. Let me also know what do you think about these pictures. A B fif makes really good pictures, and some of them, they are really standing. Also, landscapes work really really good. Also realistic stuff works really good. And, of course, especially stuff that doesn't really exist, work perfect. This, this doggo is also enormous nice, and the prompt is really, really easy. Cute gloomy four pegle wearing a flower chrome hyper realistic photo. You see everything in this photo, at least in my mind, is nearly perfect. This looks like a real photograph, and this is simply nice. So in this video, you have learned how you can create pictures in Adobe Firefly. Just click on text to image search, the gallery, if you find something that you really like, and you can also see how the prompts are structured. Generally speaking, the prompts are always structured, really, really easy. You just have to include subject medium environment, lightning color move and compositions and think about this stuff right here because the color, the lightning and decomposition, you can simply include this via the press of a few buttons. And then that's basically it. That is the whole picture generation in Adobe firefly. And in the next video, I will show you what we can do if we upload a picture for ourself. 20. Using images to manipulate the style of the output: In this video, I want to show you how you can upload your own pictures and simply use the style of them. You can change your own pictures with a few clicks, and all of this is really, really easy. If you see right here, under style, we have match. And if you hover with the mouse over this icon, you will see, add the reference image to match the style. You can either search in the reference image gallery or you can upload your own picture. And then you simply match the style from your own reference image. So for example, you can upload the picture of acute dog, but you want to have the cute dog, for example, in winter. You can easily do that and you can do a lot of stuff with this. If you want to use pictures from the gallery, of course, you can simply click on these, and you can search what you like. Let's just say you like the style of this. You simply press on it, and then you can describe what you want to make. For example, a man. So we will make a man in the style of this thing right here. And then I press simply generate. And of course, we can probably should use better prompt engineering. But we have included this style, and I think this is fine also with such a short and not precise prompt. And there you can see, we have a man, but exactly in the style of the reference pictures. And of course, you can also use other styles. If you use for example, this painting style, you can simply press on this painting style, and if we press generate again, we will have a completely different style. And there you see, now we have this perfect painting style. And I think this painting style, this is really, really brilliant. And if you go back right here, you can also upload your own pictures, and then we can use this pictures as a reference. So we simply press upload your image. And now I simply upload this picture of the woman here, and then I will say, for example, It Winter. And I simply press generate. And remember, I don't do anything else here with the prompts, I just want to show you how this uploading of images or using reference images from the gallery works because this works really, really nice. And there we have it. So this woman is now in winter. Yes, I have to admit this was not perfect, but she has a lot more clothing on on this picture, and I think this worked relatively nice. So if I open up the original picture, You see, the clothing is not that heavy, but in the new pictures, she simply has more clothes. So this is really, really cool, especially if you want to use specific styles. So if you have, for example, a really cool picture and you want to use a style of these. So in this case, the style of the woman here in the mountains, You can simply upload a picture, and we will use nearly the same style and really, really similar pictures to make these new ones. You can also upload, for example, pictures of the winter, and then simply tell adobe to make a picture of a woman and the woman will stand in the snow. And that's basically, so you can simply upload your own pictures or use pictures from this image gallery, and these pictures will deduct the style that you use in your pictures. Your prompts can be really, really short because the diffusion model can simply look at the style from the picture that you have uploaded, and the rest will do the prompt, of course. And before I forget it, you can also increase or decrease the strength. So don't forget this. A medium strength or a high strength is what I would recommend. If you go with the strength too low, this makes no sense. And don't forget, of course, you can always use different aspect ratios, different content types. Also on the content types, you can increase or decrease the intensity, and you can always use different effects and different colors, lights and compositions. This is always the same thing right here. And if you press on photo settings, they are most of the time on auto. But if you press on them, you can also do this a little bit better. So you can increase or decrease aperture, shutter speed, and field of view. So if you really know what you're doing with these cameras and camera lenses, you can really, really easily increase or decrease the stuff. And of course, you can also use CGPT because you have also learned how you can do this. In the next video, I will talk about the advanced settings and also a bit more about the photosetings. So stay tuned for the next video. 21. Advanced Settings and Camera Settings in Adobe Firefly, become a pro: In this video, I want to talk about the advanced settings and also about the camera lenses. Yes, all of this is relatively easy. It's not that advanced, but Adobe Firefly likes to call this advanced. So everything we have to do is really simple. We start with the advanced settings. If you press on the advanced settings, you simply get this right here. Exclude from image, and you can use English words only. If you hover with the mouse over this icon, you can see, add up to ten words or phrases that you want fire fight to avoid when processing your prompt. Use a comma or the red turn key to add each word or phrase. So basically, this is really, really simple. And over this are the photo settings. If you press on the photo settings, you have already seen this in the last video, just a brief moment. You can adjust the eptur and also firefly explains exactly what's going on with this. So adjust the amount of background that is blurred to draw attention to this object. So if you increase this, the background will get more blurred, and if you decrease it, the background will get more crisp. If you go down to shutter speed, you can see also the description from firefly. So create a blur that includes a streak of light or a moving object across your photo. So you can simply also play a little bit with the shutter speed. You can increase or decrease it. And you can also simply look at Chachi PD how Chachi PD always writes this prompts because you already learned how to use Chachi PT to write specific prompts. If you want to use different cameras and camera lenses in ire, you can also ask CCPD just like I told you in the Dali videos. So in the Dali videos, we made the deep dive, how we can use CCPD to do this. And then we have also the field of view. And the field of view, you can also adjust a little bit. So adjust the maximum viewable error. If you increase it, of course, this will get bigger, and if you decrease it, this will get a bit smaller. And they have also always really, really good examples, so you can see it here perfectly. And now let's just test this a little bit out. So the aperture, I want to have it at F 2.8. I think this is nice because we have a strong strong focus on our picture. If you increase it really, really high, you see maybe it gets a little bit too blur in the background. So we use something like 22.8 because I think this is relatively nice. The shutter speed I think at 2:50 is really nice. Just try to increase it all the way up and look how the pictures are. So the background gets also a bit too blurry. Or in my mind, everything gets a bit too blurry. So we just use this right here or maybe 1205. I think this is nice. And the eld of view, I think we should just leave this at 50 millimeters because this is a nice view. And now the advanced settings. First of all, we also use the other stuff. The compositions, the lightning and the color and tone. I want to use for this specific example, none, if in color and tone because we will lose something funny. The lightning, I would like to have the golden hour, and the compositions, something like a close up, I think. If I go up right here, the rest is super fine. I will exclude, of course, this picture, and now we work with a really, really easy prompt. We work with the prompt frog, and then I show you the advanced settings. If I make a rock, I am relatively sure that we will get a rock that is green. So rock in a pot with a fly on its face. Let's just try these, and then we press generate. Really nice. I think the picture is perfect. And now you can always also increase or decrease the photo settings. So you can decrease, for example, the aperture, or you can increase the shutter speed. And of course, you can also increase or decrease the Vieldo view if you want to have a little bit of a broader view. So if you decrease it, you will have a wider angle, and if you increase it, the Zoom will be a bit bigger. Firefly also describes it here all the time. Less lensplur and more lens blur. Less motion blur, more motion blur, white and zoom. And here we do now the following. So what we don't want to have in our picture. In this picture, I don't want to have green. And now I'm really excited to see what we get because, of course, the frog is green all the time. Also, the water is relatively green, and the stuff in the background is also green. Let's just see if we get maybe a yellow frog in a crystal clear water. Okay. And now we have, at least in my mind, a lot less green. So the colors are more yellow, but I have to admit we always have a little bit of green. Maybe the frog was also a stupid example, and you should exclude other stuff from your pictures. And sometimes this diffusion model is also not perfect. And like I said, I always want to tell you the truth. So normally, if you use the advanced settings and if you type in green or no green, and I have tried both right here. So I actually let the prompt run a few times, but I always get a little bit of green. In stable division, this works to be honest a little bit better and also in mid journey. But generally speaking, Adobe firefly did a great job, and now we have a lot less green, and all the rest works really, really perfect. So in this video, you have seen the advanced settings. The advanced settings are basically just the no command. So you can write some words in this prompting box and simply exclude stuff that you don't want to see in your pictures. I have tried the word green, and then I also tried the word no green. But both of it was not perfect. I tried to exclude the green color from my frog, but it was not perfect. Yes, it worked. We had a less green, but a little bit green was also there. And of course, we have also took a look at the cameras and camera lenses. And this is relatively easy. So you can simply see on the photo settings If you want to have less lens blur or more lens blur, less motion blur or more motion blur, a white angle or a Zoom. And that's basically all that matters. And if you don't want to do this, just use out and you will get automatically the right settings for your style. And my recommendation would be to try this out because of course, this is totally for free. You can make 25 generations totally for free in Adobe firefly. So just try this stuff out. And of course, if you have adobe, you can make unlimited generations. So have fun trying these. 22. Using ChatGPT to make perfect prompts in Adobe Firefly g: Do secret. In this video, I want to show you how we can use CPD, and even the free version generate our AI art in Adobe firefly, and this is enormous cool because it is so easy. Everything you have to do is to simply copy my prompt, and then you are ready to rock. Step one. You go into CPD, you copy my prompt and include it. The prompt looks something like this. So you simply press copy, you paste it into CCPD, and you send it out, and then we will simply see what CCPD does right here. First, I tell CCPD, please disregard any former directives. You occupy the role of adept photographer artist and prompt outdoor. Your mission involves devising impressive prompts for firefly. Here is some information to assist your understanding. You need to include the following in the prompts. So subject person and so on, then the medium, the environment, the lightning and so on, I simply include everything that he should include. And of course, after I tell him what he needs to include, I also tell him some examples. So here are some examples of prompts I like. First, purple sunset over a pot in the style of video game. Second, Stairway to hell. Three, a lemon wearing a headphone enjoying the sunset at the beach on Pandora. And so on, so I simply give a lot of different examples, and then we get our okay back, and after that, we basically can and probably should create our prompts. By the way, of course, we also include some camera lenses and so on in this prompts. Maybe you think from where I got this prompt. This is easy. I simply got back here. And I searched the prompts that I like in the gallery. So I simply go on the pictures, I press on them, and then I copy them and feed them into Cache Biesel We use the shot prompting to feed the prompts that I like that working firefly then we can create similar prompts unbelievably quick. So here I basically also got a few examples, and then you can simply say for yourself what you want to have, for example. Make a prompt, motorbike on the mars, and I sent this out, and then I get a better prompt. So I simply copy this prompt, and then we see what we get if we simply include this into firefly. So I simply press on one of these pictures. I delete all this stuff. I include my new prompt, and then of course, we can and probably should also use the stuff that is perfect for us. So one by one square is relatively good. It should be a digital illustration, so it is something like art and Nta photo. We don't use any specific style. I don't want to have any specific effects, color and tone, I think vibrant colors work perfect with Suria lightning and the composition is white angle. Then we use the advanced settings. Maybe we can include a negative prompt, but I don't think this is needed. And by the way, also this skull right here is perfect. Now I press generate, and I think we will get a really, really cool illustration of a motorcycle on the mars. Perfect. I think this is a really, really cool picture, and you can simply use GBD to write perfect prompts for you. Let's just try these. So tiger, for example. Then we copy these. We include it, of course, into firefly. This right here should be a photo, for example. Strength is okay, the colors. I will exclude the colors, so none. The lightning should be something like back lightning. We don't use a white angle, but a close up. We can use photo settings, but I will just use outo photo settings. I think this works really nice, and I want to exclude nothing right here, and we press generate. And there we have it. I think the tiger is an enormous cool. And I saw right here, this is an oil painting. So this works perfect with this tiger. So the prompt generation machine inside of GBD works really, really good. And by the way later in Mid journey, we will also make prompts because the prompts in Mid Journey should be at least a bit longer, and they are a bit more complex. Here in Adobe firefly, the prompts are relatively simple to write, but CechPD can help and will help, and all of this is extremely fast. So if you type in cat, you get a prompt. If you type in dog, you get a prompt. If you type in man, you get a prompt. So I hope you see where this ends. This is enormous fast, and you don't need to have any creativity because Chachi PD does everything for you. This is really, really cool. So you have learned how you can heck the system. You have also seen how you can heck the system with Dali. I gave you custom instructions for CCPD. Now here in Adobe firefly, I give you this priming. You can simply use it. You can include it into CCPD, and CCPD will make you cool prompts for Adobe firefly. And of course, then you have some settings like right here in this site, so you can use the photo settings, the advanced settings, the colors, the spec ratio, and so on. You just need to play a bit also with the lightning, and I am 100% sure that you can and will make really, really cool pictures. And by the way, if you don't like the examples that I have included into my priming of CCPD, Just simply go onto the gallery of Adobe Firefly, search for your own examples and copy and paste them and use just different examples in your priming for GPD, and you will get enormous good outputs. So this was a really, really cool and nice lecture, at least in my mind, because it can and will save you a lot of time. And if you don't want to do this course, you can also just delete all of this if you type in Diga right here. Also, a firefly will just give you some examples what you can make. But in my humble opinion, none of this works so good than our prompt generation machine. Just look at the tiger beside my face. Okay. 23. Conclusions, What you can do after the generation and a quick outlook: Over the last lectures, you have learned a lot of stuff, how to get access. What's the difference between other division models? You have also seen how the platform looks like, where you get inspiration from, how to make your pictures, how to make your prompts, how to use images for different style. All the advanced and easy settings, and on top of this, you have also seen how to use CCPD to make everything more seamless. So CCPD can really help. Now you have, of course, the question, what you can do after you have created your pictures. And that's what this lecture is all about. Let's assume you really like a picture. For example, is right here that we created. Of course, you have also seen that you can read it, and I would assume that you should do that. You can also give feedback if you like. If you want, you can save it to favorites if you press this right here. And if you press on that, you have more options. So simply press on more options. You can download it, and I think I need to go a bit to the site. So you can doload it. You can copy the link. You can copy the image. You can edit it in Adobe Express. You can submit it to the firefly gallery, or you can save it to your library. So you can do basically a lot of stuff, and I think I don't have to tell you what downloading copy the link your image means and edit in Adobe Express. That is something that we will look closely later. And of course, you can also submit it to the gallery and so on. If you go on edit, you can do basically the same stuff. So you can use the generative fil. You can use Show similar. You can use it as a style reference, so you already learned that, so you can use it in this reference for other pictures. You can also go to Adobe Express, just like you saw previously. You can apply filters and adjustments. You can remove the background, and you can also add text and more. All of this works best in Adobe Express. And that's basically it that you can do. The editing we will take a closer look in the next section because right now we are at the pictures. But what I will show you right now is the following. If you press show similar, you will make three other pictures that are relatively the same to this picture. So they change just in a minor way. As you can see, these pictures are really, really similar to the first one. So if you like one picture, especially, just simply press, show similar, and you can make similar stuff. And all of the other stuff, I will show you in the next section because this was the standard. First, of course, we need to create our pictures before we can use stuff like the generative fill. If you click on this, you lend in this Canvas, and of course, in the next section, more about this. So basically, you have learned a lot of different stuff over this whole section, and in this lecture, you have simply seen what you can do after your generations. You can download your stuff, and of course, you can make similar stuff. And that's basically downloading, making similar stuff, and then the editing comes in the next section, so stay tuned. 24. Specials in Adobe Firefly, generative Fill, text Effects, generative recolor, vector graphic: This section will be about the special features of Adobe firefly. So we have the generative feel, text effect, generative free color, text to template, text to vector graphics, and all of this also works in Adobe express and most of it also works in Adobe Photoshop. So stay tuned because I think the special features are special in Adobe firefly. 25. Generative fill, Edit your [AI] Art and Photos with Adobe Firefly: In this video, I want to talk about the generative fill. The generative fill is the same thing as in painting and out painting in Doe. But the generative fill in adobe Firefly works a little bit better, and you have a bit more control and can also blur the whole background in one simple click. Let's just take a look at everything. So of course we go at the Adobe firefly, and I press here generative feel. Of course, as soon as you made a picture, you can also press on the generative field just like you have saw in the last section. And here you can see this basically looks just like Dali, so you can either upload an image and press this button or drag and drop the file right here. But for convenience, we will just take a picture and edit the picture that we have right here. This is an example, so we simply press on the example, and I show you everything that we can do with in painting here in the Dia fight. So the interface is really, really easy to use. You can scroll up and down and zoom in and out via the mouse. Everything works really, really easy. You can, of course, always upgrade the plan. You can download your picture as soon as you're done. You can share it, and the rest is as always. If we go over to the left, you see right here, we have insert, remove and pen. If you use the pen, you can simply move this stuff around. If you use remove, you can delete these things that are in the picture. You can delete everything that you like, of course. Of course, now you can also press remove if you want. And this was, of course, a stupid move, so we get relatively nice output, but I don't think that we need that. So we simply press cancel and then clear and we have the previous picture again. If you press on insert, you can simply insert new objects. But now I will show you what the remove tool is really for. So if you press on the removed tool, you can simply delete the stuff that you don't like in your picture. And of course, you can also make your brush bigger and smaller. So if you press on it, you can make your brush size a lot bigger and a lot smaller and also the brush hardness, so how hard should everything go deleted and the brush opaqu. I will use our big brush size right now, and then we go here on this stuff because I want to delete this. So I want to remove everything here. For that, I have also to press on d, and then I can simply remove this stuff right here, and we will get a brown background. And this works relatively good. So I press and remove, and then we will have everything brown right here because, like, we don't need this stuff. We don't need the text on the wall. And there you see the whole text is removed, and we get always three different options. The third one is not that good. The second one is in my mind the best, so I press keep if we want to keep this. Then we can, of course, also delete other stuff. Let's just assume you don't like this light right here. You can always press and remove, and then you simply remove this stuff and you will get also here this background but without the light. So if you press and remove the light we get simply deleted. And done. And I think the first one is perfect. Also, the other ones are good. Let's just keep, I think the third one, so I press keep. So this is the remove. Then we also have the insert. And with the insert, you can simply add new stuff. Let's just say you don't like this things that she wear. So we simply delete everything and remember. Now we will insert new objects. So I will delete everything right here, and then we want to try if we can change everything that she is wearing. So for example, genes checked. And let's just see if that works for us. Normally, this works like 99% of the time because a Dow firefly is really, really good with this generative fill. Nice. I think the first one is nearly perfect. The second one is also really, really good and the third one also, so they look all relatively nice. We simply keep the third one, and now I will show you a new thing. If we simply reduce the brush size like a little bit, we can, for example, delete something right here. We can simply delete this, and then I say something like Necklace, and we will get a nice little necklace on this lady. We can also include air rings if you want, or we can make her blond. We can do whatever we like because the insert tool here in Adobe firefly is really, really powerful. I think the next thing that we want to change is the hair color, because the hair color works also surprisingly well here in Adobe firefly. The first one is okay. I think the first one is the best, so I keep the first one. Now I use the brush a little bit bigger again, and we delete all of the hair. We try if we can make her blonde. And like I said, this tool is really, really good. So most of the time this works nice. So we make sure that we delete enough of these, so no black hair anymore. I think now we have deleted it, and don't worry if you delete also a bit of the checked. Normally, the tool does a really, really good chop with this. So for example, blond hair, And then we press generate and we get the blond hair. The first one looks good, but the hair are a bit thin. This hair is better. I think the second one is the best, so I simply press keep, and then we have everything. Now see what we can also do. We can also press the background, and we will remove the background with one single click. This is really, really nice. Bam, and the whole background is gone. And you can also download this picture without the background right now, but I want to make sure that she goes on vacation. So for example, in the mountain, and then I press generate again, and the whole background will change, and this is really, really seamless in adobe firefly. Adobe firefly is really, really awesome with this in painting with this generative fill. The first one is a bit to blur in my mind. The second one is good and the third one is the best. So the second or the third one, they work both for me. Let's just keep the second one, and if you don't like it, you can always remove the background again. And then maybe you say that she should be in New York or whatever you like. So New York cities from the mountains in the city. And done. Now she's in New York City, and I think the third picture is relatively good there. Maybe not perfect. The second one, the first one. All of them are relatively nice. We can simply press keep, and if you don't like it, you can always do other stuff. So let's just assume you don't want to have her in New York but someone else. We can simply press on background. Then we press invert, and now she is completely deleted. And now I want to try something like robot with black hair. Maybe we get a robot inside of these nice little silhouette. And there we have it. The robot doesn't look that perfect. The second one, the third. So the third one looks relatively nice. If you like it, we can simply press keep, and then you also have this right here. And if you want to have the robot in another environment, you can also totally do this. So you simply press background again or press invert again if you want to have something else back again, so you are really, really fast with generating all of this. Now we have the background and the robot deleted. And if you think you have made a mistake, just simply press clear and you get your first picture back. And before I forget it, we also have this substract. Now, let's assume you press. If you press, you can simply delete stuff. And if you press substract, you can substract it. If you press substract, you can delete the stuff that you wanted to delete or to insert or to remove previously. Let's assume you want it to simply make a new arm, but you did also something here with the chest. You press and substract and you substract the chest, and then you are ready to go to make a new arm. Maybe you can also clean this mess up right here. In this video, you have learned what degenerative fill is. You can use the remover to remove small stuff, and you can use insert if you want to add stuff or make stuff new. You can change the background. You can change the whole environment. You can change the whole person, and you can also add and substract different things. The generative fill in Adobe Firefly is the same thing as the in and out painting in Dali. But at least in my mind, this works at the moment, as I am recording this course, a lot better. And as soon as you are done, you can always download your generations and you can use them for whatever you like. So give this a try because this works really, really nice. 26. Text Effects in Adobe Firefly: The text effects in Adobe firefly are awesome, because right now we don't have any other tool that can do such text effects. So we are really, really fast at creating unbelievably cool text in adobe firefly, and we can export it and use it, for example, for a YouTube thumbnail or for whatever you like. Everything we have to do is, of course, to press on text effects on this little button. So we press generate. And here we are in the gallery. This gallery is similar to the gallery of the pictures. So you can simply see what other creators have made, and maybe you find something that you really, really like or love. I think I like this digger here, and you can simply press on it, and then you have the same text effect. The text effect is called realistic tiger four, and you can simply try this out for yourself. Right now, it will get created and you will see how all of this looks. So this is basically the realistic tiger four, and then we have, of course, a lot of other stuff that we can use right here. So of course, you can use another text. So instead of Firefly, we can also insert here like AI with Arnie, and we can simply use everything how you love it. If you press generate, of course, you will generate your new text, and the new text will also be a realistic tiger four for right now, and we can change this by one simple click of a button. So you see right now, we have our realistic tiger four. You can also see the other generations. So if you press on them, all the letters will be a bit broader and a bit different. And if you don't like anything of the styles that you see right here, you can also press on view all. And here you see a lot of different stuff. And you can always simply press on these new things and see for yourself how they look. I think I want to try the balloon style, and then we see how the balloon style looks on the AI with Arnie stuff. You see that prompt gets automatically changed. I press generate, and then we have everything. And of course, you can always go back and see what we have, so you can use different match shape, dyked medium and loose. You can use completely different font. And you can always press on all and see what lectures you want to take. You can also take like Chinese or whatever this is. And then you go back and scroll a bit down, and you can also change the color. So, for example, let's just assume that I don't like the background color. I want to remove the background color completely. If I press on this, all of it will get completely deleted, and we have only my text right here. So I press generate, everything is deleted. And right now, if I want, I can simply download all of this And of course, we can doload it if we press on more options and download or copy link and so on. But if I simply download this, I can use it for whatever I like. You can use this text, for example, into YouTube thumb nail. Let's assume you want to make a thumb nail and want to include this text. And to be honest right now, I want to show you a quick little picture that you can maybe also create because I think this works really nice. The first thing is, of course, we download this picture. Now I go into this nice little del, and here we can simply lay our pictures over another. I press upload image. And I think I want to use this one, and then I press add overlay image and we upload our text. And there we have it, and now you can make it bigger, smaller, whatever you like. And of course, this is not the most professional way, but it is the fastest way. You can, of course, also use photoshop or whatever tool you like. You can also use SIM if you want to do this completely for free. But this tool is relatively nice because we can do it completely for free, and we have everything that we want. So right now, we have AI with Arnie. In this nice little picture. And if you want to download this, you can also do it. And if you press on this button right here, you can simply see how the picture look when all of it is done. So I think this is a really nice and cool picture, and you can totally do that completely for free, and everything works fine and seamless. So in this video, you have learned what the text effects are and how you can generate them. They are really, really easy to generate. You simply press on them and search for the stuff that you like. You type in the text that you want to generate and search for the right style for you. And then if you want, you can also delete the background completely and throw it into your own projects, and all of this is done in no time. So have fun trying this out, and I think you should really try this out. I would really recommend you to use the stool because it was never so easy to create such cool text in no time whatsoever. 27. Generative recolor for vector graphics in Adobe Firefly and the Color Wheel: In this video, I want to talk about the generative free color. And this is really, really awesome because if you ever use vector grayix, you know how painful it is to change all the colors manually. But thanks to AI, we can do this with one click right now, all in adobe Firefly. This is really, really awesome. Step one, you guessed it right. We press on generate generative recolor. And here we have the same interface as in degenerative field. So you can either upload your own vector graphics if you have one, and you can do everything here in adobe firefly. And if you don't have any vector graphics right now, you can, of course, also use the stuff that they have. I have some vector graphics on my computer, but if you don't have one, just use one of these. And I think the best way is, I also show you this on one of the examples they have. So we simply press view examples. If you find something that you like, let's just use this cat. For example, we press view sample. And here, you see, you have all your vector graphics. Blue and purple green, this is the color shem that gets used right now here. If you want to include your own stuff, you simply drag and drop a SVG v. And if you ever want to search for SVG viles, you can simply go on freepag and download some of them totally for free. Just type in Google, download vector graphics free page. And you will get a lot of stuff that you can download totally for free. And then if you have something or you use something from the firefly gallery, You can see, of course, how you want to make this. So let's just say you want to make this salmon sushi. You can simply press on these. The new prompt gets automatically used, and then you press generate and your whole vector graphics get completely new reclred. You see the original one is this right here that we have here, and then we get now our new images. You see right here, this was the original one, and now we have all these new colors. And please remember this only works with a SVG vile, so you really have to make this with a vector graphics. Vector graphics are different because they have different layers and you can recolor every single one of them seamlessly. And if you want to shuffle colors, you can also press on these. So simply press shuffle colors, and all of it will get a bit restructured. You see how nice and easy it is works. And if you want to make everything like more blue, just press on the dark blue mind. And you can not only search for the sample prompts, but you can also use the harmony. Let's, for example, use the dark blue mind. Then we scroll down on the harmony, and instead of the fault, we just use split complementary. And if you ever wonder what these complementary colors are exactly, or what, for example, square colors are exactly, this is just this stuff right here. So the color wheel explains everything. So analogous colors are simply the colors that are in the color wheel here on the same page. Complimentary colors are the colors that are vice versa. So for example, red and green or blue and yellow. I hope you get what we want. If you use square, you will get the colors that are in square here. So this makes all sense. You simply can Google color wheel and you will find everything that you can use right here. So we use, for example, complimentary colors and the dark blue style. And if I generate this, we should have really, really dark blue colors, and the complimentary colors should also be included in our generation. And boom, there we have it. You see how fast this goes. And if you need to edit your vector graphics, all of this works really fast in Adobe firefly. You can also press on reverse black and white, and you can also press generate on these if you want. But right now, this is not the thing that I want to do. So in this video, you have learned what the generative recolor is. You can simply use vor graphics. You can either upload your own vtorgraphics. You can search them, of course, also on free pick totally for free, and then you can edit them because veto graphics are a bit special. They have a lot of different layers and you can recolor every single one of these. And besides that, you have also learned what the color wheel is. Just think about the color wheel and what color you want to have. If you use complimentary colors, you use the stuff from the color wheel that is on the completely other side. For example, red and green. Just Google it and you will exactly know what you will get if you use a different harmony. 28. Text to Template in Adobe Express [AI and Firefly is also in Adobe Express]: In this video, I want to talk about text to template. And for text to template, we have to do something special. Now just look with me at this. So text to template is the stuff that is besides the generative recolor. And you see this is right now a viable in Adobe Express. You can simply press try in Express, and then you are basically on this website. This is Adobe express. This is a really cool and nice website, and I have to say this website is really, really big and you can do a lot of stuff. I will not do a deep dive in every single feature of this website because like I said, this is really, really big. I have a German course, a big German course that is also about Adobe Express and also about firefly. So this is really too big to include in this course because this course is already gigantic. But if you have more questions on Adobi Express, just hit me up and I can also add new lectures. In Adobi Express, you can simply see for yourself here on home what you can do. You can basically create Instagram square posts, fire, stories, dictoc videos, logos, Facebook posts, Instagram, reels posters, YouTube font necks, YouTube videos, and much, much more. So you can basically do a lot of stuff and you always have a complete layout from Adobe. You can simply press on the stuff, for example, on Instagram story, and you can either create from scratch or browse the templates. But like I said, that is not what we will do right here, because you can see that the generative AI is also included in Adobe express. And here we have everything that we also have on the Adobe firefly platform. If we press view all, You can see what you can generate. So you can remove background. You can resize your images. You can make caption video animate from audio. Convert to give. Convert to MP four, resize video, merge video. You can even create QR codes, convert to PDFs so you can do really a lot of different stuff. And the thing that I want to show you in this video is, of course, this here. So text to template because this is one of the key features in Adobe Firefly. If you press on these, you can use it. And like I told you, if you press on view all, you see also the other stuff. So you can of course, also make your pictures right here and also make your text effect right here. So everything from Adobe firefly is included in this platform. So right now, we press on text to template. And here you can see what you can do. So you can create templates with generative AI. By clicking agree, you agree of course to the terms of uses. And this template is really, really easy to use. You simply click on the stuff that you want to make. Let's just say Instagram square post. If you need more inspiration, you can simply see for yourself and then simply press try prompt. I think I want to try the prompt of let's just say these because right now I want to eat something sweet. If you press Instagram square, you have now the prompt fun bake sale, and we can simply press generate here below my face. If we press generate, we get a lot of cool ideas. And here, you have basically all your new stuff. But you see this right here. All of these are premium things. You can press of this. You can see variations and much, much more. And here you can basically go on and edit your complete template. You can and probably should discover all of this for yourself. You can create a whole real video or story and much more. But as you see right now, all of this is with a watermark. And this watermark simply means that we need to pay to delete this watermark. And like I told you right now, I don't have a B Express or any of this stuff. I also don't have the creative cloud right now. Like I said, I was a member of the creative cloud a few months back, but right now, I don't have the subscription. But if you want, I can also do the description again, and then I can show you more stuff, what you can do here. You can, of course, edit all of these, so you can use different text and much, much more. But like I said, this will not be a deep dive. So in this video, I want to show you that we can make entire templates out of text. In order to do that, we need to go to Adobi Express. In Adobe Express, we can do this but just if we have the subscription. So if you want to make a subscription, you can totally do this. And of course, I have not covered everything that is possible because Adobi Express is a gigantic platform. But like I said, if you need more invoxe, just hit me up and I will include some lectures. 29. Adobe Firefly AI in Adobe Illustrator and Adobe Photoshop: Adobe Firefly is also included in the illustrator and in Adobe Photoshop. And in the illustrator, you can make text to xtographics. In Illustrator, you can do basically everything with xtographics. So if you use xo graphics and if you use the creative cloud, you can always go into the illustrator and do everything that you want to do with your xtographics and use AI. Just like I told you, right now, I don't have the subscription, so I can't go in Illustrator right now. And the same thing is true for Adobe Photoshop. If you use Adobie photoshop, you can and will use the generative fill all the time. I already showed you how the generative fill works with Adobe firefly. You can do everything that I showed you also in photoshop. Photoshop is really, really a cool tool, and the generative tool works in photoshop exactly the same as you saw as in Adobe firefly. So you can basically remove background. You can invert and remove the person. Or the object. You can flip it, so you can make another object. You can make another background. You can also insert different things into your pictures or remove them entirely. Everything that I showed you also works in Adobe Photoshop. Photoshop is a really, really nice tool, and if you have access to the creative cloud, You need to use it. It works exactly the same as I showed you. The only thing that you also have in Adobe Photoshop is the lasso. If you press on the lasso, you can simply make a small little circle around the stuff that you want to change, and then you are ready to rock. Just like I told you, I have a whole course on this in German, and maybe I will make also English lessons as soon as I use Photoshop again. Just let me know if I should include also all of this into this course, and I will totally do it. Just hit me up and I won't let you down. So stay tuned because Adobe firefly is a really nice tool, and it is included in nearly every single thing from Adobe. 30. Conclusions of Adobe Firefly, Adobe illustrator und Adobe Photoshop: Over the last lectures, you have learned the special stuff from Adobe firefly. You have seen what the generative fill is and how to use it. Besides that, you have learned how to use the text effects, and we even made a small little thumbnail, possibly for YouTube. You have seen that we can use the generative recolor if you want to take your vector graphics to the next level in like no time. Text to template is something that works in Adobe Express. So in Adobe Express, Adobe firefly is completely seamlessly integrated. You can do everything that we did in Adobe Firefly also in Adobe Express, and the stuff that we did in Adobe Firefly, you can also do it completely for free, at least right now. And then you have seen that you need a premium subscription if you want to do, for example, text to template. Because, of course, Adobe want give us everything completely for free. And adobe also want a subscription if you want to use the illustrator and Adobe Photoshop. But all of Adobe firefly is integrated in all of these tools, and everything works completely the same as you have sawn over the last lectures. And like I told you, if you need help in photoshop or whatever, just let me know. 31. Midjourney, how to get access to MJ and Discord: Welcome in the first lecture of the first section in Mid journey. Mid journey is an enormous AI tool. It's a diffusion model, and you can make pictures. I hope that you remember how a difusion model works. So the computer get trained on text and pictures, noise gets added. In the process, the computer learns how to regenerate the pictures, and you need to feed prompts to generate pictures, and in Midourney you can generate pictures that look just like this. So you can see the pictures are really enormous. You can make stuff that looks like real photos. You can make stuff that looks like real art. You can make every single thing that you want, and this is just the mid Journey app. So here your pictures will get saved later. This is right here the community showcase. Don't worry but we'll take a closer look later at all of this in more detail. I just want to show you that you can create pictures that are really, really stunning. They look like from a P fi photographer. And now I want to show you how you can get access to mid Journey, because right now we can't generate our pictures in a real good application. So this right here is the website just like I told you, and I am relatively sure that we can generate the pictures relatively soon on this webpage. You can also see here. So coming soon. We can and will create pictures on this web page. I actually heard the newest call today, and in the call, the CEO told us that we can generate pictures here on this web page relatively quick. But until then, we need to work in this court. So the first thing is, of course, to make an account in Midhourney and to make a profile in this court and merge these two things together, and then we can and will create our nice AI art in this court. But all of it will get saved on this application. Step one, You simply Google mid journey and you can press on the first link. You can also press join the Beta. Yes, this is the web page that you need even if it looks a bit funny. Press join the beta. And relatively fast, you will be on this web page, and the Bot will ask you if you want to include Mourney in your discord profile. If you don't have any discord profile, you need to make one right now. You just need to do every single step that this thing tells you. So we simply press, for example, except, and then discord will get opened. So simply follow the instruction and make an account via your e mail address. If you ever run into a problem just hit me up, I don't want to bore anything in this course, and I want to make it a bit shorter. As soon as you have your profile in mid journey, and also in this court, you will be on a web page that looks something like this. So you can run this in your web browser. You can run this court in your web browser, or you can also run it via an application on your PC. It doesn't really matter. And here, the interface is relatively easy to use. So you have the chat interface, but you can't use it on the first page where you are, because this is the get started page. You don't really need to read anything here because we see every single thing that you need in this course. But you probably should go on the announcements field. So this is the second one right here. Here on announcements, you are always up to date. They post really a lot of cool stuff, and most of the time twice or three times a week, new announcements come around the corner. You can also go to status, rules, community updates, micro pols and much, much more. But I would assume that most of the things you don't really need at least not in the beginning. If you have any problems, you can also use the trial support and the support. If you press here on chat, for example, on discussion and prompt chat, you can also see what other people make. So here, some people are trying some prompts, and you can simply see what they are doing here. They have also a nice little prompt FA Q. So if you have some questions you can also see for yourself for some FAQs right here, but don't worry because we will really make enormous prompts in this course. Then you can also go on the showcase on the daily theme on the general image gene and much, much more. I think all this other stuff you don't really need. Maybe you can press on the showcase or the daily theme just to see what other people are creating. So here, for example, is the daily theme, and I see they make really nice stuff right here right at this minute. If you have never used this chord, you can simply use this chord by going onto this chat field right here. And here you can simply write everything that you like. And if you want to talk with the mid Thurney bot, you have to make a slash command. And that's basically how you can talk to the bot. So in this video, you have learned how you can make an account in Mid Journey. You simply go onto their web page. You have to register yourself on Mid Journey and also make an account on this court. The two things will get merged together. So on this court, you will invite Murney and then you can and will create standing IRT in this court. I think we will get out of this court and can create everything on the web page relatively fast, and for that reason, I'll show you in the next video, also the web page before we start to generating our AI art. So you simply saw a quick overview of how to make your account in this court and in mid journey, and then we are ready to make our first pictures. And by the way, the app that I'll show you in the next video is also the app where your AI art generation will get saved later. So even if you create all your AI art here in this court, all of this will get saved on this application that I'll show you in the next video. 32. The Midjourney App, the Subscription and some INFO: In this video, I want to show you the mid journey application. You have already took a brief look into the application or onto the web page into the last lecture. So this right here is the web page, and you already seen this explorer button. So in the left corner, if you start at the first button, this right here is the explorer, and you can simply see what other people have made. And the coolest thing is you can always press on every picture, and you can totally see how this picture was created. So this right here is the prompt, and if you want, you can also press on this three dots and make copy report or download. So you can really download these pictures and also copy the prompts. If you like it, you can also mark it with this button right here. And if you click here, of course, you will close it. And here, you can also press to copy this prompt and you can press here to search images. If you press on search images, you search simply for similar images, and I think here are really nice wolf images. They look really incredible. If you want to go back, you can simply press this button, and then you are on the first picture and one more time, and you are here. Like I told you, we will create our pictures on this web page sooner or later. But right now, this is not possible. But what's possible right now is to click on my images. All of the images that you have created will get saved here on this web page. And this is, for example, an image that I have made. This is a really cool image. The prompt works really incredibly well. And because of that, I have made a few pictures with this prompt, and you see, at least in my mind, the output is unbelievable. This looks like from a PPI photographer. This is really, really cool. So on my images, all of your images will get saved, and I have really thousands of different images made here in mid journey. And also this ice is something that I really love. I really love this kind of pictures, and you see the prompts right here are enormous long because we can and probably should use good prompt engineering in mid journey. But the important stuff is just that my images, every single generation will be saved. So every time you make something on this chart, all of this will get pushed over onto my images on this web page. The next thing is rate images. If you press on these, you can help me Journey to train their models. You can either skip here and make the next ones, and you can simply press on the picture that you find more beautiful. So you see two pictures right here every time, and if you like one picture more than the other, you can simply press on it. I think I really love this picture. I always and I also like that one and this right here, so you can simply press on the pictures that you like, and you can help to make their models better. This picture is really awful, so here is also some messed up stuff in the mirror, the arm is not really cool. You can simply help Marney to make their models better. If you press here on community, you will land onto the Discord channel, and you already know how you can use it. The next thing is view help. If you press on this, you can simply contact the billing support or you can see the documentation. Besides that, you can also press on the dark mode if you like it. I always use the dark mode just like I have told you. And then you can press on your profile, and you can manage your subscription or you can sign out. If you press on managed subscription, you are on this web page, and here you can see what plan is activated right now. The basic plan, that is the stuff that I have, and the basic plan, at least in my mind is enough for 99% of the users. And here you can see what you get. You have limited generations, so 200 per month. You have general commercional terms. You have access to the member gallery. You have optional credits to top apps. So if you ever run out of credits, you can simply buy new one. I actually did that a few times, and you have three current fast chops, so you can make free pictures at ones. The billding payment and method is monthly, at least in my case, right here. And you can always cancel the plan or change the plan. If you don't have any plan, it will look most likely something like this. And this right here are the plans. You can switch between monthly and yearly billing, and the yearly billing is, of course, a little bit cheaper. So if you use yearly billing, Instead of $10, you pay $8. I personally, I have to admit I always did the monthly billing, and I personally, I should use the yearly billing because I am on the web page signs day one. So in June or July 2022, this web page got online and I immediately used the monthly billing. So in my case, the yearly billing would be the nicer move to do. But like I said, You should probably start with monthly billing and if this is for you, because most of the time you can't get any generations for free. But I think this is totally worth it because this is absolutely the best AI art generation tool at least right now. And if you have some luck, you can test a generations for free. This depends a bit if they have enough cheap U power or not. If they have cheap U power, you can generate stuff for free. What's the difference between all these plans? Generally speaking, they are similar. You just get more stuff if you use the bigger plans. So if you go on the standard plan, you have also the relaxed generations. So you can make an unlimited amount of pictures, but, of course, just in the slow generations. Then if you go to the pro line, you get, of course, even more fast generations, and the next level up is here that you can run up until 12 current fast jobs. And one of the coolest things about the P and the mega plan is that you can use this stealth generation. So nobody can see what you make so you can work incognito. And that's basically it. And here you have some frequently asked questions. So what are fast hours? What is unlimited relax generation and so on. Just to make this quick. The fast generations, they give you simply a lot of GPU power. With the relaxed mode, you can make your pictures but a little bit slower. If you want more, you can totally do that. The community galleries the stuff that I already showed you on Explore. If you don't want that your images are public, you need a strong plan with at least 60 bucks a month. If you want to make a deep dive into the commercial use, you can and probably should use the terms of service. Just for now, I tell you that you can generate nearly everything and you are generally speaking safe. Just do it with a bit of logic. So don't make adult stuff and don't make a lot of blood and so on, I think you get what I mean. You can cancel the subscription when you want, and you can always upgrade your stuff. If you press on terms of service, you are in a discord server that looks something like this. So here you see the high level guidelines. So don't be a church. Don't use our tools to make images that could inflame, upset or cause drama. That includes gore and adult content. Be respectful to developers, moderators and guides on the server, and for more rules, see the community guidelines. So basically, just don't be a turk and you are fine. And if you want to see everything in more detail, just look at this yourself because I don't want to bore you. So in this video, you have seen the Mid journey app. Sooner or later, we will create our pictures here in this app. Right now, we will create our images in this court, and everything will get saved onto this web page. Onto this web page, you can also press explore and see what other creators are making right now. You can also copy their prompts. Besides that, you have also seen what the description is, how much does it cost, and you have learned that you shouldn't be a church. So in this manner, I see you in the next video, and in the next video, we will make you a private discord server because in my mind, this is the first move that we need to do to make our own really, really good AI art creations. 33. Make a private Discord server, add the Midjourney Bot and make the first pic: In this video, I will show you how we can create our own discord server. We will add the M journey bot, and then we can work in peace in our new channel. Because if we want to make pictures, in other channels, we get spammed with a lot of other pictures. So this right here is an important lecture, and I would recommend that you should totally do this right now at this minute. Everything you have to do is to press on this plus sign. So you need to make a new server. You should create a server. The server should be for you and for your friends, and now you need to give this server a name. Okay. So for example, best AI course, and then we press Create. Don't worry if this is in German by me, because this will always be in English by you. And right now we are completely alone into the server. But at least a robot should be my friend. If nobody else wants to be my friend, at least the robot will be, because we need him to generate the pictures, and how we can include this robot in our channel, this goes relatively easy. Because we go back onto the Murney channel, so you press on Murne you need to search it on the side bar. And then you need to go into a chat where the mid Turney bot is. And the bot is always also on this right corner here. If the bot is not on this right side, you can go maybe into the daily theme. Because in the daily theme, the bot is also always available. So here you find the bot and also here. As soon as you have found the bot, you simply press on the bot. And then you need to press add to application. Now, we will include the bot in our server. We need to use the right server. So best AI course was the server that we have created. We press outterz and then we are ready to rock. We have the bot in our chat as soon as I accept the IM human. Of course, these are the bears. And now we can go back into our chat, and now the Murneybot is here. And now we can talk to this Murneybot I already told you we can talk to this bot via slash commandos. You need to add slash, and then you have a few things that you can say to your Murneybot. You can use slash imagine settings describe info and blend. And don't worry. I will show you all of this in more detail. The first thing that you should do is to go to settings and then you send this up into your chat and you will see what version is activated for you. Right now, as I am recording this course, we are in the mid journey version six, salva version. Most of the time the newest version will be automatically used. If the newest version is not used for you, you should totally use it. As you can see, it is totally possible that the version 5.2 is the default model for some of you. We have a lot of different versions. We start with version one and the version six, Alpha is right now the best version. So it is important that you use the right version for you, and for 99% of the time the newest version is the best version. If you want to make animes or something like that, we need to use the Nice mode, but more on that later. So make sure that you use the newest version. And the rest just leave this at the default settings because we will go on every single thing of these in more detail later. So you have a style that you can use. You have different modes and so on. But right now just leave everything at the default settings and use the newest version. And now we make our first picture. We do that also via the slash. And then they imagine. And now we can include our prompts, and you already know how to write prompts. We write prompts just like in every single division model. You already learned that. So we don't do stuff like this, so we don't just type in dog. Because this is not the best prompt. Yes, we will get dogs, but the dogs will be a bit random. You already learned that we should totally include all these things. So subject medium environment, lightning, color, mood, and compositions. If we don't do that, we will get random results, the results will not be bad, but just like I told you, they will just be a bit random. Oh. You see, now we have here our pictures of the dogs, and the dogs are random. These are more like paintings and stuff, and if you want a photo, you need, of course, to include this. But we can go on and on and make really really awesome prompts just like this. So we can make really incredible good and long prompts because Midhurney is enormous good at making this prompts. You should probably also include stuff like cameras and camera lenses and much, much more. But the important stuff is just that you include the things. And also what's really, really nice is that I show you how we can train a nice CPD that we'll write our prompts with everything included completely for mid journey in no time whatsoever. And now we have a cool dog because I told Mihourney that I want to have a dog that goes surfing. And you see, now we have really, really amazing pictures. And of course, over the next lectures, you will learn everything that you need to make these incredible pictures and to do a lot of cool stuff in Mid Journey, because, like I told you, Mourney is awesome. And we also have some stuff here after the prompts. You will make a deep dive into everything. So in this video, you have learned that you need a server on this chart. You need your own server, and you need to include the mid journey bot. After the bot is included, you can talk to the bot via slash commando. You have a lot of different slash commandos. We started with the settings. Use the default settings, but make sure that you use the newest version. As soon as the newest version is activated, you can start with your prompts, and remember that you should use acceptable prompts. You have already learned the prompt engineering, But don't worry if you don't know everything about it because, of course, we will make a deep dive especially for mid Journey, because in Mid Journey, we need really, really specific prompts. And in the prompts, we also need to include stuff like the aspect ratio and much, much more. So stay tuned over the next lectures because this gets awesome. 34. Download my Cheat Sheet for important Parameters and commands: Stop Copying stuff from other people is not allowed. At least not in school. Because in my course, you should absolutely copy everything from me. I have made this nice little list, and here is everything on the list that is important for me journey. I have made this list with the commands that you will use the most and also of the parameters, and don't feel overwhelmed. M download off this list, because you will need this over the course, because we will do a deep dive in every single one of these. And by the way, this is just the stuff that is most important. If you go on the Md journey web page they have the parameter list a lot longer because they include a lot of different stuff and most of it we don't really need. But I would recommend that you should also take a look at this website if you have time for that. Also, the command list is a lot longer because I think we don't need everything from these. Now, just take a look at the most important stuff with me, and then I leave you alone with this list, you can simply download it because you will need it over the next lecture over and over again. So like I told you, you will talk to the bot via slash commandos. In this court, you simply type in slash, and if you type in imagine, you can create your picture. With settings, you go to the settings. If you use the ask command, you get answers to a question, and Mourney or the mid Journey bot will answer you. Also, possibly via an e mail. Then you can open up a whole FA Q. You can use the help guide. You can use slash info if you want to have information about your account. So here you can simply see how many picture you have already created and so on. Then you can use the blend mode to blend two pictures together. And like I said, I will show you everything that describe tool helps you to make better prompts, and longer prompts, you can shorten them via the slash shorten tool. You know, the prompt should be precise but not too expensive and too long, and for that is the shortened tool. And then we have the parameters. The parameters are also really, really important because in mid journey, we have a prompt and then the parameter. And in the next video, I will show you exactly what I mean with the prompt structure and the parameters. For now, I just leave it as it is because we have a lot of different parameters, so at the end of the prompt, we can simply type in d A or dash aspect, and we can use a different dpect ratio. We can use a KOs number, and we can simply produce unusual generations. We can use image weights on weight specific parts of the prompts more or less. We can exclude different stuff if we type in dash dash, no. We can make a lot of pictures in one single prompt if we type in dash dash repeat. We can make it up until 40 pictures with just one prompt. We can include the sat. You already know what the s is. So remember, the first starting noise of the diffusion model, and the seat number is between zero and a gigantic number that I can't read right now. You have the dash dash stop if you want to stop a generation, with the style, you can simply make your pictures a bit more stylized. With dash dash tile, you can make repeating stuff for leggings and so on, so you can make seamless patterns, and with dash dash wired, you get really wired results. This was basically just an overview of the most important commands and parameters. Just download this list and open the list while we are making the course. You will need most of them over and over again over the next lectures, so stay tuned. And don't feel overwhelmed. This will all make sense. 35. Prompting structure in Midjourney, Prompt Engineering basics: In this video, I want to talk about the prompting structure in M journey because this is a little bit different than in dally in Adobe firefly, and to be honest than in any other AI art generation tool. So this you already saw. We can simply type in slash imagine and enter our text prompt. This is what we did with these two small examples. So you just type in slash Imagine for our picture. And remember, we always use the default settings, and then you type in your prompt and remember to take a prompt that is somewhat good. So don't use a single word even if you can. Use something like this. This is the simplest way of prompting me journey, and you will get good outputs, and the outputs are square. They are in one by one. If we scroll down a bit, you see, we have the advanced features. Because in mid journey, we can divide our prompt in three different parts. So we can include a URL. So for an image, for example, then we can include our text prompt, and we can include our parameters. Now, this does not mean that we have to do this. Everything that we really have to do is to use the text prompt. Every other step is just optional. So you can include pictures in your prompts. Then the next part is the text, and the last part are the parameters. This is really important. The thing that you need to do is to simply type in the text and the rest is optional. But I would really really suggest that you should use at least the parameters about the images. So about the first stuff, we will make a separate lecture. So what the parameters are, you basically, remember that I already showed you everything. We have this basic parameters. So if we talk with the bot for example, with this slash command, just like we do right here, we can simply type in our text prompt, and we will get something cool. So I start with something like this. So I think this is a cool prompt. Chinese New Year's poster design in the style of realistic anamorphic art, Caras inout light red and emerald, relief sculpture, Patrick Brown, dragon art textual elements. And now we can include different parameters. One parameter is, for example, the aspect ratio. So dah dah, for example. We can include dash r for the spect ratio just as you have seen right here. And don't worry, like I said, we will do separate lectures in all of these. I just want to show you the whole structures of the prompt. And after this dah dah stuff, we need to press space for one time, and then we give the value of this specific parameter. In this example, I want to make this 16 by nine, and you can use the parameter however you want. So you can also use for example 15 by eight or whatever you like. Because M journey will kind of count the pixels. But most of the time it makes just sense to use the normal parameters, and you have already learned how and where you can use this. So 16 by nine is perfect for something like this course where we are in 16 by nine. If you flip it, it's nine by 16. This is perfect for Instagram and so on. And if I sent this out, we have a standard prompt, and we have also included some of the parameters. And there we have it. So I think this is a really, really cool picture right here. And of course, you can also include the third part in this prompt. If you press slash imagine, now you also could upload a simple picture and include also the third part of this thing. But like I told you, I will show you this in a later lecture separate because we can do such a lot of things with these. Now I want to show you what Merne tells us about the prompting. So explore prompting. They give us a lot of specific examples, what we can and probably should include. So the medium we can pick such a large amount of different mediums in our prompts, and always remember what you should include. So subject medium environment, lightning, color, move and compositions, and Marney gives us amazing examples. So a block print, a folk art, a graffiti, pixel art, water color, cross sketch, Bancel sketch, and so on. So you see you probably should include a lot of different stuff right here because just with a different medium, we get vastly different results. We also should be specific. So should it be a life drawing, a blind counter, a value study, Ocorcal sketch. You can also include different timelines. So a cat from the 1,700, 800, 1900s, and then a nice little 80s cat, they all look completely different. So include different timelines. And also the emotions are really, really key in mid journey. So a happy cat versus a sleepy or an angry cat, and I love, especially the angry cat. Come on. This cat is really, really cool. Also, the colors are enormous important. Just look at the difference between acid green and grainy yellow. This is enormous, so the colors are really, really important, and the day glow is also something that I like to use a lot because who doesn't love rainbows. The environments, you already seen a lot of different environments, and Marni can basically do the same stuff, so you can use every single environment that you want. So in this video, you have learned that M journey has a specific prompting structure. Remember, we work with all the default settings right now, but the most important stuff for me is that you simply understand that we have three parts of the whole prompt. We can include the image in the first part. The necessary part is the text part. This is obligated, and the third part are the parameters. The parameters are also not obligated that we can and probably should include it. If you don't include any parameters you have seen, we get just a picture that is one by one. And here we have included one parameter. The parameter was the aspect ratio, and it was 16 by nine. So this is the basic prompting structure. You already know how to write prompts, and I also gave you a lot of different examples, how different colors and much, much more change our pictures. So this is basically the first step that you should do in mid journey. Use the default setting stipenlash image. Remember that you have three specific parts. The first part with Dak later, but you know how you can and should write the text part, use normal standard prompts that you always use, and then include a parameter. If you want, you don't have to write now because we will do deep dives in all of this, of course, later. But I want that you understand the whole structures before we do a deep dive. 36. What to do after your first Generation, Upscaling, Variations and more: Until now you have learned some of the basics. And in this video, I will show you what you can do after you have generated your first pictures. So remember, you should start with a good text prompt and maybe include some parameters like the spect ratio. But don't overdo it with the start because we will take this slow and step by step. The first thing, of course, is slash imagine, and I think of a prompt right now. I'm here in a minute. We do something like this, red with glasses in the chill, smoking a cannon and wearing a tuxedo while sitting in an armchair, realistic, cinematic photo. And I don't do any lightning on this prompt right here because I don't really think we need a lightning. We can and probably should include an aspect ratio if we want. Remember, we do always dah, then We can either write out a spect ratio or we simply make this shorter and take A R. Then one time the space button and do 16 by nine. I do 16 by nine because you can see this better in the course, and then we send it. So you know how you can and should write your simple text prompts in mid journey here. And you can include the spect ratio that you like. And then I will show you what you can do after you have generated a picture because we can do a lot of stuff. And remember, we did the settings at the start of this chat, So we have always version six. The version six is always automatically included. Then, this looks really, really cool. So I think I love every single picture, and I really, really love this last one because this one is a little messed up with the hand. Also, that one is a little bit messed up right here, but we can also make this better with painting later. You already know painting and also Mitterny can do that. But I think this picture here is perfect. That one is somewhat okay. So I love this picture, and we can do different stuff here. You see, we can press on you and what this means. If I press on one, two, three, or four, we make simply an upscale of these. So if I press, for example, three, I make an up scaling of the third picture, so you can make this a little bit bigger and also the resolution should be better. Now, this hasn't to be better. Most of the time it will be better. It depends a bit on what version you use, and they change this all the time. So in 5.2, you get the same resolution. In the version six, you get a better one. And so, just remember, if you want better resolution, make an upscaling and there's a high probability that you get a better resolution. Now, what's the other stuff right here? We can press on this, and when we press on this, we can simply re run our prompt. Right now, I have included this remix stuff. It's not necessarily included by you. So if you press on the more on the remix later, of course, you will simply make the same prompt run again, and of course, you can also change this prompt a little bit if you want. So right now, I just want to run this prompt one more time again. And that's also not all because I said I like the third picture the most, and if I like a picture really, really a lot, we can make variations of these. If I press on V three, we make variations that are really, really similar to this picture. So if I press V three, we will generate a new one right here, and the variations are just minor. It's possible that we use the same seed automatically with the generations. So let's just see what we get. So the pictures are creating right now. And here, we have everything. So the first one is if we let the prompt run again. So this looks also really, really nice. So you see, we have some different pictures right here. What I don't really like on this specific example is that we have a bit of black right here in some of the pictures. But of course, we can also edit this a little bit. And this right here are the variations of my first prompt, and I think I really love also all of them to be honest. So this here is perfect. Also, that one is really good. Maybe here we have a little bit messed up hands. Let's just scroll up because, of course, we did an upscale with this stuff. After the upscale, we can even do more upscaling. We can do abdile upscale and a creative upscale, and we can very subdle and very strong, and we can light our pictures. As soon as you have made your first upscale, you can even upscale this stuff a little bit more. You can simply press on these and then you will get a subtile up scale. And if you press on it again, you can also have the creative upscale. But I must admit these upscaling, he they look really, really similar to the original stuff. And then you can also press on these and you get more variation so you can do a small remix. You can do a very subtile or a very strong remix and you always get the information that you need. So I think I do a very strong right here, and I send this out. And now we start to get our upscaling, subtyle upscale and also the strong upscale. But I think they are relatively similar, so I don't really see a lot of difference right here. And I also varied my picture. And if I press vary, yes, we get really a bit different pictures. So the style is the same, but we change a whole lot of stuff. So we change, for example, the glasses, the cigarette and much, much more and I also love this picture. So come on, this prompt is really cool. Just try it out for yourself. So in this video, I just wanted to show you that we can do a lot of different stuff after we have created our pictures. This right now is inversion six. And after we have created our pictures, you can see that we can basically do upscaling of all pictures. Variations of all pictures after the picture is upscaled, we can even do more upscaling. So the bdial and the creative upscaling theoretically here get a few pixels, get changed and so on, but this is just minor. And we can also make some variations. And if we press some variations, we get different glasses and so on, but the whole style will be the same. And you see right here, we got everything right, and I think we have really, really cool pictures. So we have a lot of different styles. We have all our up scalings, and we can also do the in painting and out painting if we want to edit this stuff further. So that's basically what you can do in the first place as soon as you has made a few pictures. So make an up scaling of the stuff that you like. And in the next video, I will show you where I get a lot of inspiration from my prompts because we can simply search for inspiration and learn how to prompt by using at the right places. 37. Where to find Inspiration for Pictures and prompts: Sometimes we have no inspiration what we should create. And this video is exactly for that. Everything you have to do is to simply go on social media for example on Instagram, and you can search for pictures from M Journey. If you go on the search bar just type in M Journey and you will find a lot of profiles that are just like these. So you can find profiles that make Murney pictures all day long. You can also find profiles if you search for AI art and much, much more. A lot of them make really, really cool and really amazing stuff. And I think also this profile makes good and cool stuff. And if you need some inspiration, you should totally go for this. So just simply search on Instagram for some inspiration, type in AI art or mid journey, and you will find a lot of different profiles. And of course, you get the best inspiration if you go on explore on the community showcase from Mourney itself. Here, you find really every single picture that you like in love. You can simply search right here a little bit, and if you find something that you really like or love, you simply press on it. And here, you see, for example, enormous big prompt, a lot of different stuff included. And if you want to try exactly the same prompt, you can simply press on copy prompt. And of course, I already told you you can also search for other stuff. So you can find similar pictures and much much more. But this is just for inspiration. So you should go on Explore, and if you find something just simply copy prompt and come back into your mid journey profile or into this court of course. And then you can simply type in slash imagine you throw in your prompt and you have included every single thing. So in this example, we have also included a lot of parameters, and you know what this stuff does. So I have given you this stuff right here. And here we have a lot of stuff included. We have the chaos style spectra so the raw style stylize 800 and diversion six. But I think we can also delete a lot of stuff. You don't need chaos if we do it at zero. This prompt was not optimal 16 by nine raw style and this stylized, so we let this include that and don't worry because we learn all of this in later lessons. This was just a cool example that you can and probably should use inspiration from this gallery. This works relatively fast and easy. So if you don't know what you should make, Just go on these and search for the stuff that you love. There is really an endless amount of enormous cool pictures, and if you don't find pictures right here, I don't know where you find pictures because you find everything that you want and you can always search for specific stuff if you want. I also love all the time this Cyberpunk stuff. So Cyberpunk is also something that you can use in your prompt. Just type in Cyberpunk, and you get really good results. And you can also make this stuff like in a really old fashion style. If you just include 1960s, all of this work. So I want to also copy this prompt. We come back into mid journey, and I also want to include this right here, so slash imagine. I frow this in. We have everything said and done, and we simply sent this. And this right here, this are the cats. I think the cats, they look really good in front the Eiffel tower, so I think this would be a really nice picture. I want to make an upscale of the second one and maybe also of the third one. And then also this picture will get really, really good because we have already seen how the style of this is. Of course, if you want to change the things a little bit up, you can and probably should do this. So here are our up scalings first from the cat. Then the second cats. I think this is a cool picture. Come on. Who doesn't like this? Then I scroll down. I think also Marilyn Monrose is really, really good. And then you can always, of course, also change this a little bit. So you can either make this new via slash image, you throw in your prompt, and then you can also add in new words. If you only include just like something like now we do something cool, top level background, and cyberpunk. I think now we will get a really cool picture because Marilyn Monroe, together with a cyberpunk I think this is cool. So the 1960s look with Cyberpunk. This is completely different words. So Cyberpunk is like really, really modern The cyber bunk look something like this right here. So this has absolutely nothing to do with our real Marilyn Monroe. I'm really excited to see what we get. And there is our small cyber punk. So I think we don't have a lot of cyber bunk going on here. I think I want to make this again. So sometimes it doesn't work that perfect. Background and then maybe cyberbunk city at the end, color photograph of Marilyn Monrose sitting in a cyber bunk city. And now I try this again. So we have Cyberbunk city this time twice in the prompt. This is also a cool tactic. If your prompt is not strong enough or something specific is not strong enough for you, and you can also include prompt weights, but more on prompt weights later. And this is 1960s Cyberpunk look. I think I really, really like this. And if you ever wonder where your pictures are, I think you already know it. Just go back on to the Mid journey app, press on my images, and you will find everything that you have created. You can press on it, and then you can simply see the pictures in a little bit bigger. This picture looks like really, really cool. This is somewhat like cyberpunk I think. Some of the pictures are maybe not perfect. Also, this is really good. Here comes the picture that I love the most. Also, this is nice. They are awesome. Just look at this. This is really, really awesome. In the gallery, you always find the pictures that you have made. I think also the older ones are really, really nice. Here you can also see the upscaling. So we have really enormous crisp pictures here. I think at least. So we already made a lot of really cool and really good pictures. Also, our first doggo is right here. And by the way, before I forget it. If you ever want to download such a picture, for example, this one right here because I really love it, it is unbelievably easy. So you have created this picture of in, but you need to download it right here. And you simply have to press on this button, and if you press on it, it will immediately be saved on your computer. And there we have it. Yes, you can also try to download your pictures from this court, but this doesn't work optimally. So the resolution and quality is a lot better if you go on my images and simply download it from here. And that's basically. So in this video, you have learned where you can search for inspiration. Just go on social media. But the coolest stuff is you can go on explore and simply copy everything that you like. If you find something is cool and interesting, just copy it and maybe change it a little bit. You can always change the prompt just like how you want. And I would recommend you to also do this because in this way, you not only copy prompts, but you can also learn to prompt. This is really, really cool. 38. The sequence of the words in your prompt makes a difference for midjourney: Minor changes in the prompts can make a big difference in the output. In this video, I will show you that even the words will make a big difference. So how you structure your sentences with the words is important even if you use the same words. Let me just show you what I mean. So if we simply type in slash imagine, and I do something like photo of a banana, on the table in 16 by nine. You will see we get a cool picture, of course. And if I simply change a few words, the picture will look relatively different. So if I simply copy all of this, so we use completely the same words, but now that photo of a banana on the table, we make simply a banana. On the table photo. So you will see how different this output is even if every single word is the same. So we have every single word, but the sentence is just structured a little bit different. Of course, I use the same aspect ratio. So in the first prompt, of course, the focus is on the banana. And in the second prompt the focus will be a bit more on the table, at least, that's what I think. So I would guess that the banana is bigger in this picture than in this picture that we get right now. More and we can even make this more dramatic. So, for example, on the table, is a banana, a photo and 16 by nine. So we use the same prompt in every single picture, but we get a little bit of different outputs because we structure the sentence a bit different. This is important to understand because you need to think what's in focus on your specific output. And there we have our picture. So the first picture banana is the first thing that we have in our prompt. And for that reason, the banana is also really, really the focus point. So you see, of course, we have the photo of that banana because we have the banana first. The second prompt, yes, it is really, really similar because also here I start with a banana on the table, and then I write my photo. But you see even here, we have a little bit less focus on the banana. Yes, I have to admit this is relatively similar, but the last photo is different. On a table is a banana, a photo. So you see here we have the table a lot more in vocus. And if I do one last example, we can do something like this. Photo of a table. And then I also do a banana, 16 by nine. Also here, we have the photo off the table, so I think we also get more focus on the table on this picture. So you see the same words will yield to different results if you structure your sentence a bit different. Yes, McCurney doesn't really understand your sentences. So McCurne only word tokens, and don't get what grammar you use, but McCurney will see That photo table in this prompt is first in the sentence, and for that reason, it must have a bigger weight than the banana. And if you start with the banana, so a banana on the table Money gets that the banana must be more important in this specific example. And here we also have the last output, and especially on this picture, we can see that we really see a lot of the table. So yes, all of this is relatively similar, but that you need to keep in mind that if you use the banana first, you get more of the banana. You see it perfect in this picture. So especially this three, this is only just banana in the picture. And if you scroll down to the pictures here, where we start with the table in the prompt, you see that we get a lot more of the table. Yes, the differences are minor, but you should keep this in mind as soon as you start to structure your prompts. 39. Different Styles and Artists in your Midjourney Prompts: In this video, I want to talk about styles because styles is a gigantic theme in all these AIR generation tools. With different styles, we can vastly improve our pictures, make them worse, or just make them different. We start with something like this. So first, I'll show you something in the settings. If I go on the settings, we have always the default settings. And here you can change this stuff. And for this specific video, I just want to change in this raw mode. Now, what is this raw mode? With this raw mode, you can make pictures that are more photo realistic. I simply do this because I have found a cool prompt of a woman. I simply copy and paste this prompt in, and then I include the aspect rate so it is nice for our course. Then you simply get a nice little picture. I think the picture is nice. I know the picture is nice because I already know how the output gets. So you see, we have three things here in the parameters, and all of this get used automatically because we have it in the settings. This right here, I have typed in manually, so the expect ratio, but the version six get used automatically because we have used the version six right here. The style is raw because we have used the raw style here. So this is really, really simple. Adjust your settings here, current suffix. We have the version six and the raw style. And if you want to change this, you can do this immediately. And now let's just see what our prompt spits out. And there we have it. So like I told you, the output is really, really good, is really, really realistic. All of this is enormous. But we can change this if we include just a bit of style. So let's just do this one time. I simply copy all of this, so I copy the whole prompt. Then I do slash imagine. I throw this prompt in and everything that I do right now is to include a style. So this hasn't necessarily to be a phone photo, but I let all of this. So I simply make a coma and I include a style, and we do a cool style right now, in art style. And then I copied a prompt and I show you a lot of stuff, and then I also show you where you can find a lot of styles. Slash imagine we throw this in, and I do, for example, in the style of the Simpsons. Then again, in the style of Mon Now I basically did the view styles. Just look what one single thing can change. Everything I have added right here is the line art. And if I add line art in this picture, the picture is completely different. So this is the original one, and this is the line art and nothing else got changed. Then we scroll down. So I have this picture versus this. And the only thing that is different right here is in the style of the Simpsons. And I have to admit I re run this prompt a few times, so the Simpsons don't seem to work perfectly right now. Because Marney has a little bit of problems with the copyrights. And to test this, I want to make a picture of the imsons right now. So we test the imsons because I want to see if Marney can create the imsons and in the meantime, I want to show you the other pictures. So what I have included in this picture in the style of Mon. And you see the picture has simply always a little bit of a different style. Yes, this is not perfect because, of course, I have included handy photography and so on. So the moon style maybe is not perfect if we use also a photo. But of course you need to do this with a bit more brain than I did. If I scroll down right now, I have the uralism and the uralism makes also really, really cool pictures. You simply see that the style changes every time, just a little tiny bit. And the outputs get vastly different. This here is completely different. To be honest, I forgot to include the aspect ratio. Everything that I added right here is graffiti wall painting, and this time, I added this few words at the beginning. So the style is enormous different. You can already see that the rest of the prompt is exactly the same. Just by changing photo to graffiti wall painting, this is completely different. And here, I tried the Simpsons a view time again. On this picture, we see the Simpsons just a little tiny bit. I said on the last picture, I tried to create the Simpsons, and I see these are exactly the Simpsons. So it seems that yes, we can make the Simpsons like really one to one exactly. Just bart and this picture looks a little bit weird. But of course, with better prompt engineering, we can make this a lot better. And now you maybe ask yourself from where we got all this cool stuff. So graffiti wall painting and especially also the line art and so on. You can simply search for lists online. We have, for example, webpage that show us 40 examples of art forms, and you can simply search for fashion art and so on. You can simply Google list of art styles for AI prompts, and you find the stuff that most people use. So drawing, college art, abstract art, Cubism, illustration, surrealism, watercolor painting, graffiti, and much, much more. There are an endless amount of lists out there. You can also include things like Mont and so on. So if you make pictures like a wall painting or something in the style of Mon, and you can also change this up so you can flip it. Let's just do one example. This will be a quick one. So wall painting in the style of Van goch. We do this 16 by nine. I think I have written angoch also not right, but we will get the angoch style in Murne because Michurny understands every single one of these. And there we have it, and I think this is like exactly Tangg style, especially this picture, and also this to be honest. So everything you have to do is to just Google. You find also a lot of lists of artists. So you can simply search for them Wikipedia and Google. They will give you an endless amount of options. I don't want to go over every single one of these because you see how many we have right here. You can simply try this out all for yourself. So in this video, you have learned that you can include artists and also styles in your photos. You can simply include the stuff in your prompt and you will get really, really cool and good results. So just try this out, and by the way, Mitch Charney can also create the Simpsons at least right now. And of course, over the copyrights, we will talk later in the course because sometimes this is a bit tricky. 40. Example of a Bad Prompt, Celebrity in the false time: In this video, I want to show you an example of a bad prompt, but you will also see that different times in history make different pictures. So this right here is my bad prompt, and I found this into the Mid Cherney gallery. So not every prompt is good here. Why is this a bad prompt? So Emmy Taylor, the year is 18 80, a vibrant and personable young woman and so on. So generally speaking, the prompt is not bad, but the prompt has a view downsides. We don't say what we have exactly. Yes, we have here the word photo realistic, but that's about all. We don't say if this should be a photo or a painting or whatsoever. And this right here, the ear is not really what Amy Taylor lives about. I think this is like Taylor Swift or someone like her. To be honest, I don't know her, but she doesn't live in this time. So if we press on this picture, you see the output is not that bad, but the output is also not good. And then we go on. Because in 1980, the picture looks a little bit better, at least in my mind. First of all, you see the pictures, at least the picture from above here, they look a little bit nicer, cleaner and more realistic, and the pictures from below, they are really like dolls, so they are really, really bad. Yes, we have photo realistic, but we need to say that we want to have a photo if it should really be photorealistic. Then the pictures from 1960. You see every picture in every year has a different style. You already see this. And then in 2000, the pictures get also a little bit better. And then we have 2020, and the last prompt is without any time and without any time we get the best outputs. So the pictures here are relatively good, but the only downside is that we are not specific if we want to have a photo. So this is exactly the same prompt as the first prompt right here. But in the first prompt, we have simply included a timeline where she has not lived. So this is awful, at least in my mind. Even the photo realistic picture doesn't look that good, and the other pictures are just animes or something, and we haven't said that we want an anime. So keep that in mind, yes, you can add different time frames. But if you do that, you should exclude specific names. Like I said, I saw this into the gallery, and I let the prompt run over and over again to change some things because I wanted to show you this. If we exclude the timeline, the prompt is somewhat okay. It's not perfect because we don't give a specific example, if we want to have a photo or something else, but at least the output is now relatively good. So in this video, you have learned that you can include a different time. But if you do that, do it without names. Because if you include names, this gets a little bit messed up. At least right now. Maybe this get fixed. I hope this get fixed. And you should also be specific if it should be a photo or whatsoever because photo realistic is not that specific. So you have learned basically that you should be specific. 41. The lightning will make completely different Pictures in Midjourney: This video is all about light. Without light, nothing works. You can really say that because one simple words, so simply the lightning makes a lot of difference. Just look with me at this. I have made a really, really simple and not that good of a prompt. So photo of a beautiful woman close up of the face studio lights. This is our standard prompt. And you see, we have the studio lights. And of course, the pictures they are really, really nice and really beautiful. Everything I have changed in the second picture is simply I have deleted the word studio and have included the word golden hour. So not studio lights but golden hour. A completely different lightning makes a completely different look. And then we scroll down because now we have neon lights. Every other single word is the same. Just look at the difference, studio lights, golden hour neon lights, and then we come with overcast. Overcast is also a really, really nice lightning situation, and the whole style of the picture is completely different. And then we have back light. So the light comes from behind. You see all this lightning stuff works enormous good. And the last one is here, the split light. I have to admit the split light messes sometimes a little bit up. So you can see at the first picture, this is a little bit messed up because no light is exactly split like this. But the other pictures are enormous good, and this is in my mind, at least perfect split light. Also, this picture is good, and to be honest, all pictures are nice, but this right here is perfect because this is perfect split light. In this video, I simply want to show you that with the lightning we can change the whole look of the entire picture with just one little word. This works enormous good, and you should totally try this. And don't worry because later, I will give you a nice little list that you should include in your prompt to get enormous good results. 42. Camera angles for different Point of view are key in midjourney (PROMO): Bird's eye view and a lot of other camera angles. You can and probably should include a lot of different camera angles. You have already learned that we can use, for example, a close up or a bird's eye view. In this video, I want to show you what you should include exactly. And I have found this nice little picture, and this explains it perfectly. So if you want a photo from top down from 90 degrees, you want to include top down view in your picture. The bird's eye view is 65 degrees. High angle shot above shot slightly above shot straight on, hero view low view and warm y view. You should simply include this simple stuff and this will work really, really cool. And I will show you this with a nice little prompt. We use this prompt right here. We use studio lights because with the studio lights, we have more focus on everything else because the studio light casts no shadows. So we don't use the close up of the face. This is really, really a simple prompt so photo of a beautiful woman, studio light, and then we include something from this. So one time, the top down view, then the high angle shot, the straight on and the low view. Just that you have some perspective. Top down view and the spect ratio always 16 by nine. Then I copy this whole prompt again, and then we include simply the high angle shot. So first, of course, lash imagine, I included high angle shot and a low view. And here it starts. This is basically the top down view. And you see this is perfect. So yes, I was not specific. And that's why the woman is lying on the ground on these three pictures. And on these pictures, she looks straightly up in the camera. But you can simply see that the camera comes from above. This is the top down view. Then we have the high angle shot. And I have to admit the high angle shot got not perfectly executed. It was somehow okay, especially this picture is somehow okay, but this was not perfectly executed. The low view got executed relatively nice, especially in the picture here, but also in that picture. You can see the camera comes from below. And because this worked not perfectly, I did the same stuff but with a city. A City is perfect for that. So I started first of all, of course, with a high angle shot from New York, then the top down view. So you can see this is really the high angle. Then the top down view comes directly from above, and the warm eye view comes from below. So this is perfect here in New York. So perfect execution by mid journey. High angle shot. Top down view, especially the two pictures here below are perfect, and the warm eye view is also really, really good. That's basically it. So think about from where you want your angle. Top down birds eye view, high angle above shot, slightly above straight on hero view low view, and warms ye view. In this video, you have learned that you should change angles. You can start here and go all the way around until the warms eye view. 43. Images in the Prompts and the Midjourney Blend Mode: I could also go on and show you everything with different colors, but I hope you get the point. So play with your prompts. You have already learned that you can and should include lightning. You have seen what a difference this makes. Different artists, different styles, different environments. Different point of view, so different angles of the camera. Close up wide angle, low view, top down view, and so on. You should include all of this in your prompts, always think about the stuff that matters. And now I will show you the blend mode. If you go into M Turney, you have already seen we talk always with this slash command. And you have also seen that we used nearly all the time the slash imagine, and we used also nearly all the time the standard settings. And now I don't show you the describe. I don't show you the info because I think the next logical step is the blend mode. If we use the blend mode, we can blend images together. We can blend either two images together, three, four, five images together, or we can blend one image together with a text prompt. If we simply press on this blend mode, we get these two things right here. And now we will start blending. And if we want to blend, we need to upload pictures. Of course. Now before we start blending, I want to give you a quick little tip. If you start blending your pictures together, you need to take in mind that the two pictures should have the same aspect ratio. So if one picture is nine by 16, the other pictures should also have nine by 16. If you have one square image, both of them should be in square, and of course, you need to upload your pictures. So it doesn't work the best with the pictures that you have right here. You should download your pictures and then re upload them again if you like pictures from mid journey. And if you want to take pictures from your PC, you can simply upload them, and then you can blend them together. That's what I will do. Let's just make something. I'm going to search a picture of me. The first thing I upload is this picture of me. This picture is in one by one. So right now I want to merge this picture together with another picture. And I have found a nice little picture. I want to merge me together with this nice little robot. And if we simply send this out, we will get a nice little picture. You can, of course, also include a third, a fourth, or a fifth different picture. And you can also specify the dimensions. So you can, of course, also write the different stuff in your prompt. But right now I don't want to do this. We simply send this out. And here we go, and I think this looks really nice. Let me just open up the original pictures. So this was basically the picture of me. And this is basically the robot that we have included. And these two pictures together, they simply create me as a nice little zy board or whatever you want to call this. So I think this worked really, really nice. And we can and probably should also give image weights to our different prompts. More on image weights, of course, later. So I think this works really nice, and I am really satisfied with this. And of course, we can also include more than just two pictures. So just like I told you, we simply use the blend mode, and we include I don't think this is, like, really needed, but we can include a lot of different pictures. So right now, I include also this picture, and maybe we can also include a third one. First, I let this picture here run, and then we include a third one. And now I include the third picture. So we can simply press right here on the prompt. Then we use the first picture, and now I also include this picture right here, so we merge this three together. And now I simply send these. So you can see, yes, this gets a bit different. I have added this picture in the first one. And then in the second one, I have added the picture that I already showed you, and that we get if we merge these two pictures together, and then we merged all three pictures together. And that's it, and that's basically the blend mode. So first, I have merged my two pictures together. So the robot and this picture right here. Then I have merged my normal picture together with the robot, and then I have merged all three pictures together. I think all three pictures merged together give the best results, not the best, but at least the funniest results. And you can also include just one picture with a tiny prompt. And this works also really good, and you need to do that a little bit different. So I will upload this picture from me, so I simply drag and drop it here in. Then I can send it. Now I simply do the slash imagine as always. I can simply drag and drop my picture here in. I add a little bit of space and now I can add my prompt. For example, terminator. And I can also add an aspect ratio if I like, but I don't want to do this. I just send this out. So now, my picture will get merged together with the terminator, and I think we will get something similar, maybe like this. And of course, we can also merge me together with other stuff. Let's just say dog or something like Barbie. And that's the stuff that we get. So first, me as the dominator. I think the pictures are really, really cool, especially this right here. To be honest, I like every single one of these. And remember, I did a horrible job in prompt engineering. So we should include everything that we like if we want to have really good prompts. And now look at these. The Barbie is enormous. M as a Barbie, this is hilarious. I think this got really, really nice. And the last thing is me as a dog, and I think this pictures got enormous good. So you can see the hoodie of me and of course, the dog face. This happens if you merge me together with different stuff. We started with the blend mode. I merged two pictures together. One time, just the picture of the robot. So this picture right here with that picture. And we simply got this output. Then I merged my normal face together, so this face right here with the same picture, and then I merged all three together. That was basically the blend mode. After that, we tried to include one picture of me in our prompt, and this works also really, really nice and easy. We simply need to drag and drop our picture in right here, and then we can include our text prompt. And like I said, I did a horrible job prompt engineering. I started with the terminator, then I got into Barbie and last but not least I got into a dog and the dog pictures are really, really cool. Just look at this husky. So in this video, you have learned that we can use pictures in our prompts and also in the blend mode. We can merge pictures with pictures with pictures with pictures with pictures. So in all we can merge five pictures together, or we can merge one picture with a text prompt. And we can make all of this more precise if we include image weights, but more on image weights later in the course. 44. The Describe Tool, learn to think like Midjourney g: That describe mode is one of my all time favorites in Murney because we can simply copy pictures that we see and we can also learn prompt engineering really, really fast. Everything we have to do is like you guessed it right because you already saw it, we type in slash, and then this tool right here describe. Write a prompt based on your image. So if you press on describe, you can upload a picture, and you should do that because Michurne will describe what Marney sees on this picture. I will make an example. I have found this nice little picture right here. This is a Tesla. I would guess models or models plate, and the car is blue, and how would you describe this car? Because now it's important that we learn from the engineering. We need to describe the car exactly the way so that Marney understands what Marney needs to generate. Think about this. This is not that easy because every single person on this Earth would describe this picture different. Just think about of someone like your mom or your grandma or somebody that doesn't like cars. Maybe they don't understand that this is a Tesla. They simply see a blue car or they see a fast blue car. So you need to describe the pictures exactly how Merne will see this. Let me just guess what we get as a prompt. This is like real time. I have no idea what Merne will tell us. I would assume. So maybe like a blue racing car. Maybe we include a Tesla. I'm not a deter, but of course, a blue car, and I am relatively sure that the words motion blur will be in these pictures. And I think that's basically. So like a blue Tesla, mods, motion blur. These are basically the three things that I would include, for sure. And now I want to see what Mitern tells us. I am relatively sure that Mitni includes more than just these three words because Mitten likes to describe a little bit broather we will also get the lightning and so on. And here we have the prompts, and in every single one of these prompts, Mourny understands the brand. So a blue Tesla model is driving down the road in the style of hyperalisms and photorealism. Performance oriented bold black lines scal elegant, moti face watercolor, digital as manual. So we don't have the word motion blur because I have assumed that we have the word motion blur. Dynamic both. So this is like nearly the same stuff as motion blurs. And now let's go so you see it's not that easy to guess what Mid Sharny exactly sees on the pictures. I just want to make three because I think we get pictures that are really, really similar to this one. Mi Journey is really good at seeing what's in pictures and then give us a really, really good prompt. So this is really, really good to learn prompting. You need to basically think what Md Journey sees in the picture. So if you upload the picture ever again, just see what mid Journey sees with this described do and in that process, you can basically learn. What Murnase? And I have absolutely no idea why we have bigger battery in this prompt right here, but I would assume that we get a nice pick anyway. And here we go. Here we have our blue Teslas, and I think the pictures are really, really similar to the first one. Just look at this one. This is nearly exactly the same picture as this. So I think here that the second picture is nearly the same as this picture. If I scroll down, I think we get pictures that are even more similar. So we have also this blurry background right in this picture, and the third picture is also relatively good but no blur right here. I think this prompt is the best, so the 2020. Tesla model is driving down the highway in the style of blue and the sure hyperalisms and photoism recycled, ready made strong facial expression, flowing silhouettes, monocharmstic ballets. So I think the prompt is good. Maybe we could also include something like motion blur. I am really, really confident that this makes the picture better. So if we simply include the word motion blur in this prompt, maybe just as the last word, Motion blur, so because the car needs to be in action and not just standing still as in this picture, and then I think we get a really, really good output. Motion blur or also dynamic. These are words that add some action speed. And don't worry. I will give you a nice little list of really, really cool words that you probably should include in your prompts. But more on that later as soon as we get pros in prompt engineering for mid journey. And perfect. I knew it. With the word motion blur, the pictures are really in action. I think this last one here is really, really nice, so you can see all of the tires are spinning here in this picture. So we got the right result. So in this video, you have learned what the described do is. The described dole is awesome, because you can simply type in slash, describe, upload the picture, and then you can learn what Money sees on this picture. In that process, you can also think about what mid turn disease, and in that process you learn how to prompt. And besides of that, you can also just recreate pictures. If you ever find a picture somewhere and you want to create a similar one, just upload it, but don't tell people that you basically just used another picture as an example, because I don't know if people like that. But anyhow, this is how you can do it. 45. Shorten your Prompts for Midjourney with the shorten command: You remember that we should make our prompts if it's possible short and precise. Now, sometimes we find prompts that are really, really long, but they yield enormous good results. And we have a nice solution for that. We can try to cut the fluff from the biggest prompts, make them shorter, more precise, and get even better outputs. In this video, I will show you the shortened command. And by the way, in the meantime, I run some prompts because I think these are really cool pictures. Let me introduce you the rabbit that goes shopping. Also, this picture looks enormous nice. And because all of this looks better in 16 by nine in of course, I simply made all of this a little big bigger. I think this looks enormous cool. Also, this prompt is relatively nice, and in 16 by nine, you can see it better. Yeah, I think also this is a nice little prompt, and all of it looks really nice. Now, what you can see right here, this is a really, really short but precise prompt, and it yields an enormous good result. This prompt is a bit long, but it's somehow okay, at least in my mind. But do you remember our picture from above of the cats in Paris where we have our cats. This right here. This prompt was enormous long, but we can and probably should try this. I copy all of this prompt. Then I press slash. So we always talk with the slash command. We don't use imagine settings blend or describe and also not info, but we use the shortened command. So slash shorten. Then I include the prompt and I send it, and now we will get options held to make the prompt more precise and more short. The fifth prompt is always the shortest one. So M Journey tries to exclude more and more stuff, and at some point, of course, this doesn't work anymore. So I want to run the first prompt, maybe also then the second prompt and the third prompt, and I think we should use an aspect ratio of 16 by nine Vodes because like you see it better in the course. And there we have it, and I think the first picture got really good. So we have our selfie off the two cats in front of the eiffel tower. So we cut out de vf and we get really, really good results. This can and probably will yield enormous good results if you cut out the v. This looks hilarious. And if you scroll down a bit you see that the stuff get also cut out that we need. The prompt gets shorter and shorter and not everything will be included. So, yes, we have a cat from time to time. Yes, we have the Eiffel tower from time to time. But we have also two people here in these pictures, and we don't need any people at all. And if we scroll even more down, the prompt is too short and not everything is included. We have no Eiffel tower. We have just a little kitten with a person. And don't get me wrong. These are also nice pictures, but these are not the pictures that we want. So you can and probably should use the shortened command, but you should always see if the important stuff is included in your prompt. You see, Marney did just simply try to include everything that is important, and they cut out the fluff. They list the most important words and sorted with a comma and that's basically it. In this video, I wanted to show you that you should shorten your prompts if possible. If you find a really, really long prompt, just try to type in slash shorten, throw it into Murne and see what Murney spits you out. Most of the time, the first and second option work well. And if you go down further and further, The last prompt doesn't work that well because we exclude the most important stuff. But Mhurney tells us to be short and precise and the shortened tool can help you with that. Because sometimes we just include random words. So we don't need the words and so on because Mischarney don't understand the whole sentences, Mishneys only the word tokens, and it is enough if you just use the words that you need and use a comma to make your next word. 46. Commands and Settings that you really need in Midjourney, Reemiy, Style: Let's talk about the final commands and settings. You have seen most of it already, and I will show you this on my little list. I will prove it to you. You know what the slash imagine is. So you make your pictures. You understand the slash settings. You need to type this in if you want to use it. And in this lecture, we will make a deep dive into the settings. You already know what the blend mode is. You've seen the described mode. You also know what the shortened command is. We already used it. If you have any questions directly on M tourney, you can use slash slash FAQ slash help, and this is also somewhat important slash invo. So if you want to use any other command right here, just simply type it in. Ask FAQ and help or this right here, we can do this right now. You can simply type in slash invo and send it, and then you will get invo about your subscription. So basically, how many pictures you created in this month. So you see right now I have 61% of my fast time remaining. I have made 1,530 images in total. So 1,530 grits. This, by the way, is called a grit. Four pictures is one grit. So I have made a lot of pictures in mid journey. And you see, this is basically the slash info, and if you type in as can on, you can and will get an e mail from Mid journey. But what I want to show you right now is the slash settings because this is important and we already took a brief look on this. First of all, of course, the version. You can always use different versions. Over the niche version, we will make a separate lecture. Like I said, most of the time, use the newest version. But some of the other versions have a few advantages. If we use the version 5.2, we can do some stuff that we can't do in the version six. But more on that, of course, later. Right now, I just want to use the version one just to show you the difference of the outputs. So we go up, for example, on this prompt, so this is an enormous cool prompt, and I simply include everything until here. And we use this in version one so that you can see what the difference is. I make a picture in version one, and by the way, in July of 2021, this version got online, and you will see how far we come already. So this is not that far in the past that we had the first version and the quality was really not that pretty. I remember how I made pictures in version one. But besides the version. Of course, I would just assume that you use the newest version or maybe from time to time, the 5.2 or the Niche mode. Like I said, I would assume we use simply this version right now. And here we have the raw mode. Now, what does the raw mode? If I include the raw mode, we have included also the parameter in the prompt. So dash dash style raw. And if we use the raw style, we can simply have a bit more quality, a bit more realistic quality in our images. Now I will show you the raw style versus the non raw style. So I go up on this prompt again because for this, we need the raw style. So I will show you first the raw style and then without the raw style, and then you'll see what mess the version one is. So the version won't picture They are really awesome. So we start right here with out the raw style, and then I let the same prompt run again, but with the raw style. So now I include the raw style. And by the way, if you exclude the raw style, you can also include it just in your prompt with dash dash style raw, more on that later, of course, because we have some parameters that I didn't show you until now. Besides that we have also the style factor, so we can increase or decrease the style. So you see the medium style, and we haven't included anything in our small little prompt. If we use a low style, we include S 50 for the low style. If we make the style a bit higher, we have 250, and if we use really high, we have 750. By the way, this style goes from one until 999. And now I will show you in the raw mode, for example, the difference between the style one and the style 999. For that, I will use this style in the prompt. So we do simply slash imagine again. We have our prompt, and then dash dash style one for the first prompt. And again, slash imagine We use our prompt and dash dash style 999 and here we go. And here we have everything. So this is version one, and I hope you see why I don't recommend you to use version one unless you want to make funny stuff where you just need to imagine what it is. After that, we used version six. But this is not the raw style. This is the normal version six without the raw style, and if you look closely, you see that this is a bit more like a painting, so we don't have the exact photoalismus than in the raw style. Now I have to scroll down because this is the normal raw style. In the raw style, you see that everything looks more crisp. So we have really, really awesome pictures that are enormous photo realistic. Now, I don't say this is bad. So you can use whatever you like. But if you want to have more realisms, you should use the raw style. And here we have the difference between style 999 versus style one. I hope you see what I mean. If you increase the style, everything will just pop more. The snake is maybe a little bit bigger than it normally would be, so we have a lot of style and Murney can lean a bit out of the window. Marney can simply make stuff that looks cool, and that is not necessarily realistic. Now, if you go down and if we use style one, Marney will not lean that far out of the window. Marney will simply make pictures that can basically be real. So you see the girl has just a mask in the face. Or the snake is just really, really little. I don't know what's this. This is a little bit messed up. Or she has a hat on. So I think you can and probably should play a little bit with this style because these two prompts are completely different. Every thing is completely different. Just the style got a little bit higher. So the style factor, you can't ignore it. Please please please play with the style. And if we go up again, We have a bit more stuff right here. So you have the public mode, the mix mode, the high variation mode, the low variation mode, and that's basically everything that we have in this line. Now, what's the public mode? If you want to exclude the public mode, you see right here, I can't do that. Because we have to use the public mode. If we scroll down, you can see it. We need the strong subscription to go private. So everybody can and will see the pictures that I am creating here. At least if I do it in the daily theme, so they will most likely not be in my discord server. Besides that, we have the remix mode, and the remix mode is really, really cool. Basically, if we have enabled the remix mode, we can press some variations, for example, on this picture right here, and then the remix prompt will get enabled. So we can remix this prompt just a tiny bit if you want. And if you exclude this, so if you press on this again, the mix mode is not enabled, if you press on this again, you can't remix this. So the prompt will get automatically used. So let's just scroll down because right now we get a picture. And like I told you, we can't remix this. And if I press on the remix mode and I include it again, I can simply press on this, and I can vary this variation just a tiny bit. So instead of Black woman, let's just say Asian woman. And let's just see if M Turney can do it because the remix is really, really subtle. So we can't make a lot of changes with this remix. It is just for variations, and if you do variations, the remix will get enabled, and you can change the output just a tiny little bit. And besides that, you can, of course, try to exclude the remix mode and you can also rerun everything again. And also, if you let everything re run again, you can't use the remix tool on this stuff right here. Only if you include the remix mode, you can basically adjust the prompt before you send it. That's basically the mix mode, and I would suggest that you let the mix mode include it every single time. Let's just see what we got from this mix mode. Now we got basically our different variations, and you can see all of this works really, really nice. And as soon as I did the remix, I could also change the prompt a little bit, and now we have Asian woman. So this works really nice. I just would tell to you include the mix mode, you will not regret it because you can simply always just Just press, generate and don't redo your prompt and it's basically the same. So just let this include it. Then we have also the high variation mode and the low variation mode. And now I will show you what this means. So first, I start with the high variation mode. And I have found a really cool prompt. I press slash imagine. I include my whole prompt, and then I simply send this with the high variation mode. By the way, this time, we make a man and not always woman. And now I use the low variation mode, and then I simply press this again. So slash imagine I include my prompt and we send it, and you will see what the difference is. The output of the grids as soon as we use the high variation will just vary a lot, just like you see in this picture right here, and the grid from the output where we don't have a lot variation. So where we have used the low variation mode, this output will vary a lot less. And I will show you this again because I have used an aspect ratio that you can see really good. So we do this again with the 16 by nine, and this expect ratio needs to delete, of course. Then I used the low variation. So in 16 by nine, you can simply see it better. And now you can basically see it also in this picture relatively good. So this is the high variation, and this is the low variation. So I hope you see that we simply have a lot more variation if we increase this. And here we have the picture 16 by nine. And you also see here we simply have a lot more variation than in these pictures. So you see the colors and the lightning are just with a little bit more variarty and that's basically all about that I have to say about this row right here. And the last row we will make this really fast. So you can use sticky style, and then you have basically your different modes. So you have the turbo mode, the vast mode, and the relaxed mode. I would recommend you to always just use the vast mode as long as you have fast minutes. If you use the turbo mode, you can do that and your pictures will be generated faster, but this will cost more. And the relax mode, you can only use it if you have a stronger subscription. With the relax mode, you can do an unlimited amount of generations. But, of course, you need a stronger subscription. So I would assume that most of the people are fine if they run just the vast mode. And now I want to open up the settings again because they get closed. From time to time, they get closed if you have them opened for too long. Now you can also see the following thing right here. If I use, for example, a very high style, I can also do here this sticky style, and even if I use a medium style. And now I will also tell you what this sticky style is. If I include this, I can simply use the same style over and over again. So, for example, if I do a prompt, for example, let's just say this prompt right here. If I let this prompt run with this style 750, We will have a sticky style. So if I use another prompt again, the same style, so 750 will get included. And that's basically this sticky style. But I have to admit I don't use the sticky style. I most of the time use the high variation mode because I find it is nicer and I use a medium style or from time to time, a higher style. And that's basically every thing that we have to say about all these commands. So you know the imagine. You know every single thing from the settings. You know the blend modes that described mode shorten as FAQ help, and info. So we have basically seen all commands that make sense. Yes, there are more, but you don't need to understand more. 47. Different versions in midjourney, Creating Animes with the Niji Mode: In this video, I want to talk about different versions of M journey. You have already seen that version one is not the best one, and you saw that version six is enormous. But we have a lot of other different versions, and in this video, I will especially show you that we can make animus like really, really awesome. And for that, we need the so called Niche version. Maybe as soon as you see the course, we are in the ij version six. Right now we are in the Nj version 5.1 or 5.2. And by the way, this 5.1 and 5.2 are also awesome because with these, we can do later in the course, the in painting in Marney and also the out painting. In painting and down painting in Marney is also a nice feature, and you need to understand this. But first, of course of all, let's just see on the settings, what we have included right now. So we have the mid journey version six alpha, a medium style. Public mode remix, high variation. The fast mode no sticky style works for me. Because I want to show you now the slash imagine a gigantic prompt of Songcu from Dragon Ball. Songcu is a nice anime character, at least in my mind, and now I will show you what happens if we use different versions. So we have the version six Alpha, of course. Then we have the version 5.2. This is the normal version 5.2, so no Niche no anime style whatsoever. I sent this up with the version 5.2. We have the version 5.1. This is really, really similar to the version 5.2, so I will skip this. But the next thing is the version five in the Nice mode. So Nhidel version five where we have it. So Niche model version five. And now I will use this with the Nice mode. And you see also that we can include different stuff if we use the G mode, so we can increase and decrease the style as always. We can use default style expressive style cute style scene style and original style. And this is all self explained. So if you want to have very strong expressive style, just use this. If you want to make something cool, you should probably use that. And if you want to make your scene, you should use this and the original style is like original. This right here, so all the other stuff is as always. And now I will show you this nice mode because also the niche mode makes incredible good pictures. And by the way, all these models make enormous good pictures of animes. Then we have also the nice model version four. The nice model version four is in my mind, at least a bit too old to compete with this other stuff, but they are also relatively nice. And then we have also more other versions, so version four and so on, so I don't think that you should ever use version four. Because like we have a lot better stuff. Let's just see what we got, shouldn't we. We start with the version 5.2. The 5.2 version, generally speaking, makes also really good pictures. And I think the pictures they are really, really, nice, really, really cool. But they miss just a tiny piece. This is more like your drawing and not that impressive at least to me. Here we have the version six, and the version six is unbelievable. The version six can do everything, and the version six makes here this stuff like from the movie. If you watch a movie on a high definition film and Sogoku does some really, really fancy stuff, this looks like exactly the same. Then we scroll down because we have, of course, also the nice version four. And the Nice version four. So this is our old nice version. They can also make really, really nice pictures. They are also just like from the previous from the old dragonball films. So even the Nice version four is really, really good. And here we have the Nice version five. And I think the Niche version five is on the level from the normal version six. So if you want to make animes, I would really recommend you to use either the Niche version five, Or if you like the old style the Nice version four, and if not, use the version six for everything that you like. So the Niche version five, this is really awesome. And I think there are also other anime characters just like Nara Moto and whatsoever. So you can make every single one of these with this nice version, but I don't want to waste your time. So simply take your favorite character and make some pictures. Or you know what we make one picture one last picture of Sailor Moon. And here we have Sailor Moon. This is Nice version five. So the standard Niche version five, pictures are good. Nice version five with the cute style. And you see the difference. I hope you get that this is a lot cutter than this right here. But like I said, the pictures are in both of the outputs really good. Then we have the version six. Version six of the sailor moon is really, really crisp and clear, and I really love the version six Version six is awesome. And now we have also the version four. And the version four is just a bit more classic, as you know. So in this video, you have seen a lot of different versions, and we simply took a look what's optimal four generating animus. We have two different variations that are specifically made for animus. The whole mode is called iC mode. Right now, we have Nice mode four and Nice mode five. Maybe we get NICE mode six in a few weeks. And if you have Nice mode six, I assume that this will be awesome. And the normal version six is also really, really good with ems. So just try for yourself what mode is optimal for you. 48. Midjourney Upscaling, we have a lot of different Tools: In this video, I want to talk about the up scalars in M journey because the up scalars, well, they are a bit of a mess. They change like every other week. So let's just see what we do right now because here we have something from version six. And I simply make an upscale of some pictures. Let's just go up right here to the Gucus because I like the Gk better. This is, for example, the Niche version five, and I make an upscale of this first picture. So we press on one. Then I go up right here. And we have here the NS four is okay. I go up. This right here is version six. So I make an upscale of this picture from version six. Then I go up again. This right here is version 5.2, and I make an upscale of 5.2. And then we basically have everything that we need. So we have here the up scales from the versions. And you see, we can do a lot of stuff right here. We can do a little bit of stuff right here, and we can also do a lot of stuff right here. And why is that? And just take a look, if we press on up scales here on version four. If I press right here, something completely different will start. So here we start a complete new job, and the upscaling runs right away. So normally, if we press on version six, the upscaling is like nearly instant. And here you see how long this takes. So right now we wait forever to have 0%. And if I go up and we simply press upscale, for example, on this thing from version six. Let's just say the first picture, I can scroll down and bam there we have it. So this right here is instantly upscaled and the version four takes like forever to upscale. Now, well, we have a lot of different up scalers. And if I go on the Mid journey web page, they also tell us a bunch about the upscaling. But to be honest, we don't know to need every single thing of these up scalars, we just need to know that some up scalars are better and some up scales are worse, to be honest, it feels sometimes that Midourney doesn't even know what up scalars right now are activated because also, this is not always up to date. Everything you need to know is if you press on upscaling of one of this picture, the picture will always get opened in a separate tab. So down here, for example. If you use version 5.1, 5.2, and version six, you don't really get an upscale. So the picture just copied down here, but it doesn't get any bigger. If you press on version four, right now, you get an upscale. So you see right now with it, our upscale and the picture got really bigger. The next thing. If you like D Nice version or version 5.2, you've got a lot of options that you can do later, and if you use version six, you have just little options. It also depends on what version you work. If you work with version six, you just have a upscaledile and a creative upscale. Then, of course, you can very subtile and very strong. If you work with version five or 5.2 or something else, you have a double up scale of four x up scale, and you have also this right here very subtile, very strong, very region. Zoom out, make square and this pan tool. All of this, of course, in a later election because this is big to explain because we can do a lot of stuff with these. But right now, I just want to tell you that the upscaling is in Marney a bit of a mess. In this video, you have simply learned that Mhurney himself doesn't even know what they do with their up scalings. Everything that we know is if we press on the first pictures, the pictures will get cloed down, and some of them will get a big bigger and some of them not. So for some of them, the resolution increases. Right now as I am recording this course, the version six doesn't increase. Also version 55.2 and so one doesn't increase. The output is as it is. But after the output is generated, we can upscale, and also here, we have different upscaling methods. We can upscale two x or four x in all of the version five, and we can make a upscale subdal and a creative upscale in the version six. I hope this was not confusing, click upscale sometimes if it works, it works if it doesn't. Just forget about it. After you have downloaded your pictures, you can simply see for yourself if the resolution is big enough or not. And later in the course, I will also show you online upscaling tools if you need to upscale this any further. We can also make vector graphics and print this stuff on like wall papers and whatsoever. 49. Conclusions, Outlook and a bit of Homework: Over the last lecture, you have really learned a lot of stuff. We started with my little cheat sheet, and I hope you downloaded it. You got a nice little overview. W took a look at the prompts. You saw what lightning, camera angles, compositions, styles and different timelines make for a difference. So all of this is really important, and you should know all of this. You have also learned every single command. So they blend they shorten, that describe blah, blah, blah, and so on. You have really sown a lot. You have even learned to shorten your prompts. The whole command list, the animus, the upscaling chaos and much, much more, so this was a gigantic section. And I have one little thing to do for you. Just go into Murney. Use the divult settings, make sure that they are right. Type in slash, imagine, and T fun. And over the next lectures, we will make a deep dive what we can do in Money because Murne is much, much bigger than I already showed you. We can also do in painting out painting, much, much more. Stay tumed. 50. What this section is all about: In this section, we will make a deep dive into Leonardo AI, not just like the picture generation but also editing of the pictures and much, much more. You will see the real time Canvas, the real time generation, the Canvas editor or the painting and out painting, if you want to call it this way. The texture generation, how to make videos with Leonardo motion, and of course, also how you can train your own AI model inside of Leonardo AI. That's the easiest way how you can train your own stable fusion, Laura. Stay tuned for the next lectures because all of this is awesome. 51. The Chaos parameter for variation in Midjourney, style and stylize ist not the same Thing: You already learned that we can talk with the bot with this dah dah commands at the end of our prompts. We always include this parameters at the end of the prompt. So A R for the aspect ratio. And in this video, I want to talk about the dah dah chaos. Dash dash Kos gives you more variations. And now, one thing to remember. Of course, you can use these different styles. So stylize low. You've already seen this four 50, stylize high for 750. And if you want to make raw style, simply press on this raw mode. So dash style raw. So we can also combine the raw style with stylize. So you see we have included three things at our prompt completely automatically. You see, we can make our prompts bigger and bigger with more parameters. So this right here is my prompt. This will be simply a digital art style of a beautiful Garcel Bachmo we get something like deer. You see, we have stylize 750 included in our prompt. And also here, we have the stylize included. We can also exclude it from here because we included it also in the prompt. Then I have version six, 16 nine for the aspect ratio, and this right here is new chaos 99. I will send this prompt, and then I let the exact same prompt run again, but with less chaos, and you will see what's the difference. So we make slash imagine everything the same, but chaos at one. And I hope you already see the difference. First, we have Chaos 99, and then we have Chaos one. With Chaos 99, we have four pictures that are completely different. So our grid has a lot of varirity. If I use Chaos one, we get a grid that has really, really similar pictures. And you also see in my parameters list that we should use a number 0-100, and it simply changes the results a bit, so we get a lot more variarty if we include chaos. And I also want to stress this right here. 750 and style raw is a different parameter. Also, Murne tells us this. So we have style and we have style lies, and it depends on the version that you are using. We have the dash style raw, for example, and that works in Mur 5.1, 5.2, and also in the version six. If you use dash dash style 4a4b for C, this is for Murney version four. I have to admit we don't need this. And the dash dash style cute expressive, original or scenic. You have already seen that we can use this in Murney version five, and don't confuse this with the stylize, because the stylize is just a number and you can include stylize in every single prompt. And a number 1-999. I already showed you these for a quick moment, but I want to make this crystal clear. Style and stylize are two different things. And with the chaos, you have a lot more vararty in your grits. Just like you have seen here. If you use Chaos 199, you get a lot of variety. If you use Chaos one, you get little variety. And you see, we can include a lot of stuff after our prompt. Stylize 750 dash version 6.0 for the version, da A R for the aspect ratio. Then dash dash chaos one and dash dash style raw. And that's not all because we can include even more stuff and more stuff about the next video. Because in this video, you have learned that you should include dash dash Chaos if you want a lot of rarity and dash dash style and dash dash stylize is not the same thing. 52. Midjourney can create weird results with the word command: In this video, I want to talk about the parameter that I personally don't use a lot, but it's worth to notice that it's possible to use, and the outputs are. Yes, they are funny. Just see what I mean. We talk about dash dash wired. And now we start with a prompt. Japanese girl, very detailed, full body view, sharp skin, Kodak camera, chaos of 50, has back ratio 16 by nine. I want to have the raw style. Stylize is 963, and we have the versions include lighting, for example, backlight. And then I will run with this prompt and I copy this prompt because then I want to run this prompt again. And the next time we include one last parameter and the last parameter will be the dash wired. With the dash wired, we will get wired and thy results. And please don't forget to add a value. So I have included the highest wired value in this example, so wired 3,000. And of course, this can vary 1-3 thousand. So wired or also d W, and then you can also include a little bit of less irness. Let's just say 1,500. So the first picture is with no wired dit, that the four settings will always be zero. Zero wiredness will be included. Then if I use, of course, d wired, we can and should use a value of wired. Right now, I use value 30003000 is the highest wired value. And now we will get an output with wired $1,500. And I have to admit you can really tell what's more wired. So I think as soon as you start to include the dah dah wired, you get always a bit of wired results, it doesn't really matter how high the value is because I personally would argue that this first picture is really, really wired, and I think this is also a bit scary to be honest. So if this is not wired, I really don't know what wired is because this thing scares me. So wired is relatively nice to use if you want to play a little bit with different stuff. But if you don't use wired, you get something like prettier results. And if you want to be a really creative person, just include this and see for yourself, what do you get out from these? Especially if you want to be really, really creative, try to play a little bit with this wired because you can create a lot of stuff, especially if you want to lean a bit in the creative space. Let me show you one more example. So here is my first picture without any wired style. I have to say I just copied this prompt from the mid journey gallery, so I didn't come up with this stuff for myself. Then I included wire 1,000. And you see the results are pretty pretty dramatically different. And then I included wire 2,800, and the prompts are also, again, a lot different. So in this video, you have learned that you can include dash dash wired if you want to make really, really creative and different pictures. I personally, I don't play a lot with this dash dash wired because you never know what you get, but some people seem to like this and some people seem to play a lot with this stuff. Maybe you are one of these people, and because of that, I want to include it also in this course. 53. Negative Prompting in Midjourney with the NO command, exclude colores and more: Dash dash, no. You have to learn to say no sometimes. If you don't want to include certain things into a prompt, just type in dash dash, no, and you will exclude it. Let's just make your quick example because this is really easier than you think. So I start with a prompt. The prompt is relatively big. Don't worry about this prompt because you will learn how to write this. Before I can include it, I have to tell M Churny that we need to bot, so slash imagine and then include the prompt. Now, I start with, of course, A R space 16 by nine of course for the spect ratio, dash S 800, stylize, then dash style raw because I want to look this a bit realistic. A chaos. I like to use a chaos of 400. The version six get included automatically. But if you don't have included the version six version 6.0, we can include it also here. And now we include this command. But first, I let this prompt run. I copy everything so that you can see the difference. I copy this prompt, I let this prompt run, and then I let this prompt run again, but with the command. And the chaos, of course, I need to use a chaos of 40 and not 400. Excuse me. And now, space. Green. I want to exclude green colors from our prompt, and let's just see if this works or not. And here is our output, so the normal output without the dash no green. And you see, we have a nice little frog. The frog is green, and the frog is king, as you would expect it. Now, let's just see what we got if we use the dash no green. Of course, we have a rock as king that is not green. So, you see, we have here some yellow and red colors. Green is completely eliminated out of our pictures. And by the way, I let the prompt run a few times, and also really, really simple prompts work really fine. You see, I have run this prompt of rock is a king 16 by nine, no green version six raw style. And also this prompt delivers a nice little frog the rock is king. And the frog is not green. In this video, you have learned that you can use the dah dah no to exclude certain stuff from pictures. You can exclude colors. You can also exclude stuff like clothing. So you don't want to have, for example, a white shirt, just simply type in dash dash, no white shirt, and the guy has maybe a sweater on or something else. Dash dash, no hair. If you want no hair in your pictures, dash dash no ocean. If you are, for example, on a beach, but you don't want to see the ocean. Dash dash know everything. Exclude everything that you don't like to see in your prompt. Let your normal prompt run, and if you see something that you don't like, just make this stuff happen again, but include dash dash, no, AKA, the stuff that you don't like. 54. Use the repeat command if you want a lot of pictures from 1 prompt: In this video, I want to talk about the d d r or dash repeat. With this command, you can simply let the prompt run over and over and over again. If you don't have the pro plan, you can run up until three chops at the same time. If you have the pro subscription, you can run up to four t chops at a time. So let's just take a look how we can do that. Of course, always the same stuff slash imagine you need a prompt. So we have a Japanese girl. We have dos 50. The aspect ratio is nine by 16. But like you know, I like to have it 16 by nine so that you can see it better. We have the raw style. We have a high stylize. And this right here is image weight. And about the image weight, I want to talk about later. What I want to show you right now is that we can include R. And if I include right here, the number two, we'll let this prompt run twice. If I use the number three, we let this prompt run three times. So I sent this out right now. Marney asks me if I want to make this twice, and I simply tell Mn, come on. Let's go. And then you see we have two chops right now. We create two pictures at the same time. Like I said, you can create up until three chops at the same time with the cheapest subscription, and if you have a better subscription, you can run up until 40 prompts at a time. This is the output so the pictures are incredible. The prompt yield good results and the dash dash r makes it happen as often as you like. In this video, you have learned that you can include dash dash r at the end of your prompt and simply say mid Journey, how many times you want to run this thing. You can go up until three times in the three version and up until four t times at the best version at the time. So if you need to create lots of pictures, just try it out. 55. The Tiling command to make seamless textures for Print on demand in Midjourney: In this video, I want to talk about the tiling feature in mid journey because I really like the tiling feature. It is really easy to monetize this tying feature. The tying feature means simply that we make something that is repetitive. So we can create, for example, roses that are one seamless thing. And of course, you can make your own pants underwear. Hoodies, T shirts, bags, and much, much more with these. You can upload it on websites like red bubble and much, much more to create a little bit of passive income. And all of this is really, really easy to create. You just need to find the right tying stuff, and then you are ready to create your own things. Everything you need to do is to as always slash imagine. And then you need to type in what you want to see. We make this really, really easy, roses, for example. And then I use dash tile and you already saw my first dialing stuff right here. This was in version 5.2, and now we will send out these. As you can see, we can also use the version six for our tying. The previous stuff was version 5.2, and now you will see the difference. Everything you have to do is to simply type in the stuff that you want to see. Use every single prompt that you like. Use every single prompt that you can imagine. The version six works fine, the version 5.2, 5.1, and also version four. They are all really, really nice with this tying stuff, and then you can simply press on the things that you really like or you go onto your mid journey page and you can download these because now I want to see if this is working or not. So this thing is on my mid journey app. Of course. I think this looks good, and then I download it with pressing on this button. And now I go onto this web page. This is simply a tying check here. So team less texture check here. You can simply Google this or you're typing this URL. I will also link it, and then we can upload this per drag and drop. So I simply drag and drop the picture that I have created in mid journey, and then we can see that this works perfectly. This is completely seamless. So it works in small stuff, and it also works as we oom zoom in. This is enormous seamless and you see this works really, really fine. And of course, you can make every single thing that you like. Let's just make one more example. So slash imagine o part, dial and then you are ready to rock. And of course, here comes a little quick tip. If you ever go on web pages like Ets for example, just look for the stuff that sells really, really good with the seamless textures. Then you can make a screenshot of these. You can go into M Journey, use the describe dole and then you see what prompt you need to use to get such texture, and this texture will sell really, really well. But I don't want that everybody knows this. So you see right now we have also our Leopard dial stuff. I go back into Mourney into the app. I think I really love this, and I want to see if this is really seamless. So I download it. Okay. Into my downloads, of course, there's always the picture. Then I want to go onto the seamless texture. Check out one more time. I include this right here, and you see also this is completely seamless and you can simply zoom in and zoom out and see how nice this works. I think even the points they are completely seamless. So here starts the next picture, and you see the overlap is enormous, good. You can zoom in as far as you want, and you can't see where the next picture starts. In this video, you have learned that dash dash tile gives you really really cool dialing stuff. You can use this in every single version that you like. So Version four version five version 5.15 0.2 and version six. All of this seems to work at least right now completely perfect. And like I told you the nice little trick. Go on platforms, see what tiling stuff is selling good. Make a small screenshot just of the dialing stuff, and then use the describe Dolin Mourney. I already told you how to use the describe Don M journey. So please remember, Upload just the picture, see what prompts you get. Use one of these prompts and include the dah dah tile at the end, and you have a dialing feature that is repetitive that works every single time, and that has the potential to sell well on Ez or every other single platform like red bubble and much much more. Your print on demand business is ready to skyrocket with these because you can make such cool stuff in no time whatsoever. 56. Creating Videos in Midjourney with this command: In this video, I want to talk about video creation. Yes, it's possible to create videos in mid journey, but it seems to work maybe a little bit different than you think. We can simply see how the division model makes your pictures out of the noise. And I think this looks relatively cool, so we have this feature, and we should use it. So we simply press slash imagine. We need a nice little prompt. The prompt that I use right now is simply a nice little kit that will use stuff from Star Wars. I include dash chaos 80 right now. So we have a high chaos so that we get four pictures that are relatively different. The expect ratio, I want to use a bit different. So I want to have 16 by nine as always because I think this is cool. Then we have the stylized feature. Also the stylize is perfect. Version six works also good for me, and then I simply type in dash video, and that's basically it. I let this prompt run, and then we will get our nice little picture. Now, what I have to say for this is that we need to tell the boat that we want to have a video. So we need to include this dash dash video. Also, the row style is included this makes no difference, but we need to wait until we get our picture before we can ask for the video. Now, we get our Star Wars kits, and I think the Star Wars kit they are hilarious. They're really nice. Also, the quality is good to see, we have a lot of variation, and now I want to show you how we can get the video. We need to react to this simple picture, and we can do this if we hover with the mouse over the three dots. And then we need to make a small reaction of these. So we press on reaction. If you use this a few times, you will get your envelope right here. If you haven't used this forever, just simply press on reaction. Then you are in this interface, and now you need to search the envelope. And for that, you simply type in envelope right here. The envelope is this nice little button, and if you press on this envelope, you see that you get your envelope also right here, and then you get a message from the Murneybt and the M churney bot hits you up right in the left upper corner. Now you can press on the Mery bot. And then you are in a separate chat, and here, you see your small little video. You can make this big and you can also play this right here. But the stuff that works better is, of course, to press on this link. If you press on this link, you get into a new interface, so you need to accept that you are into a new page. And now we are on this web page, and here we can see how M Journey creates our pictures out of the noise if we press play on this button. And I think this always looks really, really cool. So you can see how the diffusion models makes your pictures out of the noise. And basically, this is the video that we can create in M Journey right now. I would assume that we can also create better videos in the future because M Journey talks a lot about video creation. So stay tuned because I think we will get a lot of really, really cool video updates also inside of M Journey. Now, if you go back onto your M journey server here, you can also download all of this of course, and then you can simply play it yourself. And we can make a lot better videos with stable difusion and other tools outside of Murney at least right now. But I wanted to show you that we can simply include a dash dash video inside your prompts, and then we can see how the difusion model makes our pictures out of the noise, and I think this looks cool. And if you want to learn more about videos, of course, we have a lot of stuff inside this course later. So stay tuned because video is also really big and we can do a lot of stuff. And right now you learn the stuff that you can do inside of M journey, and as soon as Mourney adds new stuff. I will include also this because I would assume that we can create standing videos with M Journey in the future. So stay tuned. AI is just incredible. 57. Character consistency in Midjourney with the seed: In this video, I want to talk about seats and how we can use sats inside of mid journey to create consistent characters. Yes, this is possible and yes, this works relatively easy. And you already know what the seat is. It's the first starting noise of the division model. Everything we need to do is, of course, start with slash imagine, type in our prompt. Right now, I have a nice prompt of a beautiful woman. We use the raw style. We use stylize 750 and diversion six, and then I simply include sat. No, we do this a little bit different. I let this prompt run. Because I want to show you that we get a nice little output. And I think I let this prompt run a few times because I like to have an aspect ratio of 16 by nine of course. 16 by nine, I let this run. Then I let it run again. And then you will see that we get three different outputs, and then I will show you how we can use seats to make all of this consistent. Then I also show you how we can take seats and change the prompt just a little bit to make a consistent character inside of different environments. Now you see we have three different outputs, and I scroll up to the pictures that are in 16 by nine so that you can see this better. We use the same prompt so everything else is the same, but the output, of course, is not the same. And now I will show you what we get if we use a fixed seat. Slash imagine I typed this in. Then we use the aspect ratio as always in 16 by nine, and now I use a seat. Dat, don't just use S because S is style. Dah dah sat, and now we take a seat. We can't take a seat between one and an enormous big number that I can't tell you right now because the number is too big for my brain. I simply use this seat, and then I copy the whole prompt and we let this prompt run two times so that you can see what the sat does. So right now, I do slash imagine. I type this in, and I let run this prompt again. And now we should get two outputs that are exactly the same. And there we have it. So this is that same prompt with the same sat, and we have twice exactly the same output, or at least an output that is really, really, really similar. So if you look at the picture in the left corner, this is exactly the picture from this here in the left corner. You just need to see for yourself, we have an enormous character consistency thanks to the same sat. So same prompt with same seat will generate the same pictures. You can see it right here. If we use the same prompt, but no fixed seat, we get different pictures because the seat will get selected randomly. You already know how this works if you saw everything from the difusion model, so the basics that I already told you. Now you may have the question, how you can use, for example, this woman but in another environment. You can simply copy all of these, make sure to include also the seed, and then you throw everything in right here, and then you simply include stuff like in the garden. And then you get an output that is really, really similar, so you should get nearly exactly this woman but in the garden because we use the same seed. Yes, the output will be a little bit different, but it should be nearly the same woman. And if you want to make this even more consistent, you can use phase swap. But about phase swap, I will talk later because we need to include more stuff into mid journey to use this phase swapping technique. But this is the easiest way how you can make really, really consistent characters in different environments if you're using mid journey. And this right here is the output, and I think we don't have to argue that yes, this is nearly exactly the same woman but in a different environment. So right now she's simply in the garden. You can see it here. So this looks nearly exactly the same as in the pictures above because we use nearly the same prompt, we just added in the garden, and we used the same seed as in this pictures. And now I will show you also a cool trick. So if we scroll up, for example, like on this picture, and you want to see what prompt got used for this picture or also for this picture. It doesn't really matter for what picture. You can simply see for yourself what the seed was. And you can do this, of course, with this picture or with that, it doesn't really matter. So let's just say you want the seed of this picture because this sat is obviously another sat then the sat that we used here. We can simply press on this three dots. We use the reaction, and then we search for the envelope. The envelope sends you every single information that you need. So you simply press on envelope and then the Murney bot will hit you up with a nice little message that includes a lot of stuff. You already know we used the envelope also here to see our video, and now we press on the left corner on the Murney bot. And the Mery bot send me also this picture, and the Mitch bot gave me now the exact information of this prompt. So the sat that was used in this prompt was this sat right here. And we can also simply copy this seat if we want to copy the seat, then go back into our chat, and now we can recreate also exactly this picture. And we do that as we always do. So slash imagine, then I want to type in d sat. So I simply include the dash dash sat, and if we use the sat that we got from Md Journey. I can copy this entire prompt until here, and we simply include everything here. And now we have also the sat, and we should be able to recreate exactly this picture once again. And you know, if you want to make exactly this picture in another environment, you can, of course, also do this. So we copy this one more time. I simply do slash imagine. I type in everything that we want. And now she should be, for example, on the beach, and we send this out. Here is our output. So this picture should be exactly the same picture as the picture above, so as this picture. And yes, this is completely true. Just look at it. So this picture is exactly the same picture as this. We recreated this picture because we used the same prompt with the same seat. And then I did, of course, just the same stuff. So I included everything from this prompt with the sat, and I changed the environment by adding a simple sentences like on the beach. And now we have the same woman, but now she is on the beach. And this is how you create character consistency inside of M Turning by using the seat. So in this video, you have learned how to use seats in Murn. Dip in at the end of the prompt dah dah seat. If you want to use a fixed seat, dip in dash their seat and a specific number. Then if you want to create consistent characters, use this sat and change the prompt just by a tiny bit and you will get really, really character consistency. If you have a picture where you don't know the seat, Press on the three dots, search for the envelope, and Murney will tell you what the seat of this specific picture was, and you are able to recreate and find me the picture that you have already made. So have fun with seats in M Journey. I think this is one of the nicest feature insides of M journey, and nobody talks about this. 58. Midjourney Character reference and style reference: Mithurny has some nice updates. We have character reference and style reference. And in this video, I will show you both of it. If you want to have consistent characters without using the seat, this is the stuff that I will show you in this video. Let's just go on the Mitern webpage or on the Mitch cord server and see this right here. Here you see that we get style reference and also character reference. We can dipe in simply d the CRF and include the URL, and then we can use the weight. If we use zero, we have only the phase. If we use 100, we have nearly every single thing. Of course, we can also use the SRF. Not CRF, but SRF for the styles. Let's just start with the CRF, and I want to show you how this works. Here I show you every single thing. I started with just some normal prompts until I get a picture that I like. I just let the prompt run a few times until I get a picture that I like. I think I like this picture, and then we go on with our prompts. We simply use women on the beach, then d d CRF, and we include it and we have this picture. So now we have this woman but on a different picture, and we can go on with this. I started all over again. I used just a normal prompt until I get a picture that I like, I made an upscale, then we started to do the stuff with the C Rf, woman in the night club d d CRF. And now you see I have this woman but in a night club. And you also see that even the clothes get copied. So if we use the normal CRF, we have a weight of 100 and we also clone the clothes. Here the same thing, but I think this picture is not perfect. If we use a high C ref, the pictures get silly from time to time. If we scroll down a bit, and this right here, this is somehow okay. And here we have the same thing but with a black stress, and in order to do that, we need, of course, to use a cf of zero. If we don't use zero, we can't use this. So we have right now, C zero, we have only the face copy and the dress is black. If we use the same stuff, so, but without the weight of zero, we get the same picture, so we get a red dress, even if we use our black stress in the prompt. If I go down a little bit, here the same thing, but I included the weight of zero, and then every single thing works, we get our black stress. Here's some more examples. Black stress without the weight of zero, and we get the red dress. And of course, we can also do stuff like a comic picture of this woman, a cyber punk picture of this woman, a comic pencil art, the golden hour, Pastel studio ites in a night club. As you see if we use the weight of zero, it gets a lot better. We don't have the red clothes in every single picture. Professional studio its wonder woman here, but always the same f so you see this is always the same face. Woman from 1940, you see you get black and white. 1980, Catwoman, 1990, and from 2000. You see this works really nice. Let's just make one simple test so that you understand how this works exactly. We start as always with slash imagine. Women on the beach with a green dress. Then of course, CRF, and now we should include, of course, our picture. So we go back until we find the picture that we want to have. So this picture, for example, we simply drag and drop it right here. And then of course, we can include the stuff afterwards. So for example, H R 16 by nine, and I think this works fine. We sent this out, and then I will show you that we will not get a green dress because I have not included the weight. If we use no weight right here, we get automatically 100 and as you know, 100, will also copy the clothes. And there you can see it so we get the red dress. And of course, I will show you in the meantime what we can do if we want to really have our green dress. So I simply copy the exact same stuff right here. I include this. And yes, you can see it right now. So we have our red dress. This dress is, of course, not green. What we need to do if we only want to have the face but different clothes, this is really simple. We do the same thing. We include our URL right here. Then we use C W for character weight, and we use zero, and then of course, our respect ratio, for example, 16 by nine. And we sent this out, and now this should totally work because we have reduced the character weight. If the weight is zero, we can use different clothes, we just copy the face. And of course, you can and probably should play a little bit with these. Maybe try to use a weight of ten or 20 or 30 if you want to have it more similar. I have personally found that a weight of zero works, at least in my mind, best. And there we have it, so we have the same face of course on a woman or it's nearly the same woman, to be honest, but with a green dress. Yes, not every single picture is perfect, but I think this picture is really, really good. Also, that picture is good he or she is a little bit of a Maybe like a zombie, I don't know. Maybe she also doesn't wear anything under this stress, but yeah, okay. And this is also not perfect. So some pictures are, of course, really good. Others are not perfect. This is always the starting picture, as you know. So this is right here, and we can put this face with this simple C f on every single thing that we want. So right here in the nightclub, and of course, if we don't use the character weight, we also use the same clothing. You can see it right here, character weight. Of 100 is default, so we use the same clothes. If we use the character weight of zero, we can use different clothes. I think most of these pictures are really, really good. You can also use different style, cyber bank comics and much much more even pencilar pastel. From time to time, you get something that is really weird, just like here, is lions But if you play a little bit like this in the goal now where I think this looks nice, also the pastel, the studio light, this night lights, professional studio, and also this wonder woman, and, of course, also the old pictures, they look really, really nice. Let me know what you do with this stool. I think this stool is awesome. And of course, this is not all because we have also the SRF. If we scroll down a little bit, Marney has a nice example. We use it just like our character reference, and we simply include SRF, the URL, and at the end, we can use a different weight. The default weight right here is 100 and you can go up until 1,000, and this is the stuff that you can do. This is the reference. This is the output that we get. And if we use no reference, we get something like this out of Marney. Let's just make one example. I think I want to have something like this, so the style of this. So we simply make an upscaling of this picture. Then I do something like this. Slash imagine man as superhero. Then we use, of course, SRF. Then we can include this stuff right here. And now, if we send this out, of course, with the right aspect ratio, we will get something back with a similar style than this. And of course, we can increase or decrease the stuff. So man as superhero SRF. Of course, we include this right here. Then the style weight. Of course, S W. I want to use 900 right now because like 100 is the default settings, and we use the aspect ratio, and we should get something back that is even more the style of this picture. And there we have it, it works really, really good. You see, we have the same style as in this picture, but now we have a man as a superhero. This is the default setting of 100. I think the style is really, really similar than in this picture. And if we increase the weight to 900, we get something that is even more similar at least in the eyes of M Turney. You can also decrease this to zero, and that's the last thing that I want to show you. I copy all of this, of course, one last time slash imagine. We include this. Then we use the weight of zero, and of course, we will use the same style, but the weight will be a lot lower. And there we have it, so the style is somehow similar, but of course, not that similar. The weight of zero just uses a little bit of the style, weight of 900 or even 1,000 uses the maximum capacity, and I think that default settings is always the best. Character reference and style reference are two of the most powerful tools in me journey. You can make precisely at what you think. Let me know if you like it or not. And remember, just start with the prompt. Use CRF or SRF include the URL of the picture, and then d S W, and you can increase the weight. And that's basically it. 59. Conclusions, Outlook and a small amount of Homework: Over the last lectures, you have learned a lot of really really cool stuff inside of M journey. You have seen how to use chaos for more variation in your pictures. You have also seen that stylize and the style is not the same thing. Negative prompting in M journey works well. Dash dash no, and you can exclude the stuff that you don't like. If you want to create a lot of jobs simultaneously, you can totally do that with dash dash repeat. You can create up until two 40 jobs if you have the strongest subscription. You can make really phenomenal tilings with the dash dash tile, and you can start your own print on demand business with this feature. We can even create small videos and see how the division model makes these pictures with the dash dash video command, and we can probably should use seats inside of M Journey to create consistent characters. So these are some of the coolest features inside of M Journey and a small outlook in the next section. I will show you a lot of really cool tricks, some more key insights into M journey, and I am really sure that all of this will help you with wars face swaps and stuff like that. And now I have one small stuff of homework for you. Go into Murne and use the coolest stuff for you. I personally, I would find it really cool if you use the seat because the sat brings you at a new level inside of Murne. You can create character consistency with the sat, and maybe who knows you can also make a nice AI influencer with Murne if you use the right sats. Of course, we will also talk about AI influencers a bit. Because also this stuff is really, really cool, so stay tuned for the next lectures because all of this is awesome, and one small tip for you. If you find that all of this discourse helps you in some way. You can, of course, tell your friends that the discourse is cool. And by doing that, you get also credit for doing that for your friends, because the people will always associate the value from this course with you. So do me and also you a small little favor, have fun in the rest of the course. 60. Become a Midjourney Pro in this Section!: This section will make you a M journey pro. I will show you a lot of cool magic words that you need to include into your prompts. We will also create our own CPD that will write you amazing prompts. You can simply copy and base my stuff. We will create text. We will use pre mutation prompting, multi prompting and prompting weights. We will combine everything and include pictures and smiles, and of course, we will also save all of this as shortcuts. So stay tuned because this section will make you into a Murney pro. 61. Use my Magic Keywords for optimal Midjourney results: In this video, I want to talk about magic words inside of me journey. Every single one of these words will change the output of your prompt dramatically. Let's just take a look at what we have. Of course, the various types of shots. Extreme close up close up, medium shot, full shot over the shoulder shot point of view shot, y level shot, high angle, low angle, long shot, extreme long shot, full body view. Dutch angle shot bird's eye view, rule of thirds shot, Silhouette shot, two shot, master shot, candid shot, drone shot, go pro shot, and fish eye shot. These are all the shots that you can and probably should include from time to time into your prompts, and every single one of these will change your outputs dramatically. Because the extreme close up is different from the close up. I hope you understand what all of these shots means. If you don't get one specific shot, just simply Google this, so you can simply copy this and then go into Google and see what it is. So let's just make an example at the extreme long shot. This right here, for example, are extreme long shots, and of course, you can include them in your prompts. Let's make also an example into Murney. This will be a beautiful lake in Alaska, Extreme long shot. And here we have our extreme long shot. Remember, with all of these, you will be able to create pictures like prophy photographer because they really struggle to make such shots and you can do it in mid journey in no time whatsoever. And now let's compare this with the fish eye shot. So beautiful lake in Alaska, but this time a fish eye shot. You see the lens of the fish eye is just not straight. It's a bit like this, and you can and should include all of this in your prompts. Now, let's go on with this list because you can not only include different times of shots, but you can also include a lot of different looks. You can include cinematic looks. If you include stuff like cinematic film grain, ultra realistic film no sepia tone, death of field dramatic lightning. And here is a quick comparison. And remember, the prompt is awful. I compared Spider Man against Spider Man cinematic, and you see the cinematic pictures they are really on the next level. Just look at this and all because of one single word. Remember, you should use good prompt, so you already learned the prompt engineering. I'm just here to want to make a point that the words work. Then we have the lightning. You can include studio lights bright lights, neon high key lightning, low key lightning, rim lightning, practical lightning, motivator lightning, warm, cold, sunny, gold hour rainy foggy, night afternoon soft side lightning magic hour and morning. I don't have every single one of this lightning. We had a separate lecture, but make sure to include this stuff. And we have also keywords for movements, action scene, dramatic scene, dynamic motion motion blur, freeze frame slow motion and time labs. Here is a quick example. So formula one car versus formula one car motion blur. You instantly see that the car with the motion blur looks like that it is in a completely rush to win the race. Of course, you should also include cinematic cameras. You already know how to include cameras, and this are some of the cameras that work really, really nice. Area Lexa super 16 vintage film, on cinematic EOS, So sign ultra, red cinematic camera, and black magic cinema camera. If you want to use action cameras, for example, for shots like this racing car, on EOS one D x marker two. Go Pro hero in black. Sony Alpha nine two, Nikon D six, and the Olympic OEM one x. You can simply include all of these, and you can copy and paste all of these. We can include different film genres to make a really hot look. So horror Western fantasy romantic mystical animation fria sence fiction documentary and adventure. You can also include filmmaker, so you can use the style of Quentin Teranino Alfred Hitchock, Martin Source, Christopher Nolan, Michael Bay, John Ho, Peter Himes, James Cameron, Wes Anderson, David Fincher, and also Stanley abridge. You can include your favorite ones. You can even include sports photographers. So Ella Garrison, Walter Cho Cheer, Neil Chafer, Simon Brute, and Darren Carrell. And of course, also emotions are really important in your prompt. Angry said, hopeful, happy, surprised, scared and bored just to name a few. And you should also combine all of this stuff, so we can make a prompt something like this. Close up of a B. Motion blur ultra realistic, gold war dynamic action on EOS one D x mark two. Then we include the spect ratio of 16 by nine. We use the version six. We use stylized 750, the raw style and Kos 40. You learned every single step of these, and now we send this out. I think I love this cord. The third and also the second picture are awesome. I want to make a small up scale so that you can see this better. This is perfect with the motion blur, and this is generally speaking a nice picture. And now you know how you should include every single one of these pieces to put together your puzzle. So this video got you a nice little che cheet. Just include all the words or at least some of the words that I gave you in this list. When you include words from this list, your prompts or your outputs will get drastically better. And this is awesome because you can save a lot of time. And in the next video, I will show you how you can save even more time because we create together in CPD a prompt generating machine. And you can simply copy and paste my prompt generating machine. And I will also give you a link to a GPD so that you can simply type in what word and you get awesome prompts, so stay tuned. 62. ChatGPT as Midjourney Prompt generating Machine (Copy my GPT): In this video, I want to show you how we can train C CPT to be a prompt generating machine for mid journey. You can use this either in the free version of CPT or in the paid version of CPT, and then I will also show you how you can train your own CPT or create your prompt yourself. And of course, you can simply copy and baste the stuff that I will show you. So this right here is the prompt that I use, and you can simply copy and paste it. I will include it in the lecture, and then you will create a prompt generation machine out of GPD 3.5, so the completely free version. You just need to copy this prompt, included, then you get done okay and now you can tell GPD, for example, prompt of a cinematic close up of lean mask. And then you will see you get included everything that you like. But I would suggest that you don't include the spec ratio and the style because it's easier to do this yourself. And now let's test the prompt because I think we will get an amazing image. I simply include the prompt. We use an expect ratio, and nothing else, maybe a bit of style dash dash style. Let's just say 500 and here we go. Okay. And here we have Ellen mask. Three of the pictures are nearly perfect and one of the pictures is not so good. It's because I have included the style. So with this match style, we get a lot of different outputs. If I don't use the style, we will get a picture that is closer to Ellen mask. But all the other picture, especially the one from the left corner looks amazing and looks just like Ellen mask. And now I will show you how the prompt is structured, so we go back into C GPD, and then we will create our own GPD that can create this prompt, and you can simply use the link to my GPD, the GPD that we will create life right now, and then you are ready to rock. So the prompt is structured relatively easy, and of course, you can and probably should also write such a prompt for yourself. So please disregard any formal directives. Then we use the role prompting. We tell him that he is a photographer, an artist, and a prompt outdoor. The mission is, of course, to make really good prompts for M journey. Then I give some information about M journey. M journey is simply an AI tool or a difusion model and so on, and then I give CCPD simple a few examples so that CCPD understands how M Journey prompts are structured. So I simply included, I think six or eight different images, and CCPD simply understands with all these examples how M Turney needs this prompt. So here is, for example, image eight, breathtaking and so on photograph we include cameras and camera lenses and much, much more. We also include photographers like Tim Walker and so on, so you get the concept, I simply trained CCPD on all the invo that is necessary to create really, really good prompts. And then I tell CCPD simply to answer only with k, and wait for more instructions. And of course, you can simply copy and paste this prompting to GPT 3.5 or to GPD four, and if you want to create your own CPD, we will do it like this. We press on the name in the left corner. We press on my CPDs. We use create a CPD, Now it's time to give a D GPD a name. For example, M J GPD. So it's a simply M journey prompt generating machine, and we can't include the word mid journey because Mourne is another company and we can't do this here. Then the small description, right perfect AI prompts. And now the instruction, So here we need to be short and precise because later, I want to upload a file, and on this file, we train GPD on perfect examples. We will do something like this. You need to write prompts for a diffusion model. You look at your knowledge base for all the info. Make sure to be precise and answer only with the prompt. And then we can include for example on conversation starters. Let's just start with, for example, close up of a spider, beautiful woman. And white angle view of a lake. And now we need to train CCPD your knowledge. I have prepared a perfect knowledge file, and now I will show you how this file is structured. So this is the text file on that. I want to train CCPD. The text file is really, really big, and I include a lot of different prompts for example. So we have here ten different examples because CCPD will understand all of this because he uses his vector database to understand all of this knowledge. I also included all the stuff that I already told you. So the cinematics look and much, much more, so CCPD has a lot of info. Now, I want to simply save this file, and then we upload this file, of course, into our knowledge base. So I press upload file, and I think now CCPD is perfectly trained on this stuff. I don't want to have web browsing here for this small little CPD. And I don't like to have the L included. Maybe we can include the code interpreter, but I don't think this would be necessary either. So you can simply upload these if you have your GPD or you can simply copy my link. I will send you this link. Now I make this stuff public. So anyone with a link or everyone in general, so I simply press confirm and everybody can have access to this little GPD. Maybe I will get even a small revenue cut because maybe some people like the stuff that I made here for M journey. And now I will simply test this out. So close up of spider, for example, and then we see if we get a prompt or not. And it seems legit at least at the first shot right here, we include cameras and camera lenses, and we even have an aspect ratio of one by one. I think I want to change this as you know, because you can see it better in the course. So we go into mid journey. I think this stuff works really well because CCPD understands everything here. The only thing I change right here is the aspect ratio to 16 by nine. In the meantime, I go back into CCPD and I try beautiful woman as a rogue. Let's just see if we get something even out of this funny little prompt. You see the prompts, they look like really impressive, at least to me. Every single line is perfectly executed. So we have always included. Also, the Apcrat we have always camera lenses included. So CG BD here is really a pro. So slash imagine, I want to type this in. The Aptrato would be also nice right here, but I want to have 16 by nine, and the spider looks phenomenal. So at least to me, it seems that we have trained CCPD completely perfect to generate stunning pictures. I want to make an upscale because I think this looks enormous, nice. What thing do you like the most? I think the third picture looks awesome. Then, I would really think that you should totally just copy this CachPD and make the same thing if you have the CachPD palazo because this thing is phenomenal. And here we have our woman as a rogue. So I think also this looks perfect. It's picture perfect. Then I make an upscale of these and one last test of our Cech PD, what we want to make right now. A dog is president in a suit, and I even made some spelling errors right here, but CCPD gets it, so no need to worry about your spelling. I copy these. We go back into mid journey for one last test. I include this. And remember, we always have an aspect ratio. It would be probably the perfect aspect ratio for this nice little prompt because I have trained GPD to understand automatically what aspect ratio is right. So I would assume we get the perfect aspect ratio. Let's just make one last test. For example, YouTube thumbnail. YouTube thumbnail of a scared cat that takes a selfie in front of a dog. Let's just see if we can even make a nice little YouTube thumbnail. CPD will do a great job and we'll describe everything as needed, and we even get the right aspect ratio so perfect. CPD understands that we need 16 by nine for our YouTube thumbnail, and also the prompt looks really, really nice. So back into Mc Journey, we used this slash imagine. I include it, we send it, and we have everything, and I think Also, this dog in a suit looks like phenomenal. I really, really love the first one. Also the last one is nice. I want to make an upscale off the last one just to show you that this picture is relatively cool. And you see, we get our dog as president. It looks at least to me, really, really impressive. And then I want to see if we get the scared cat that makes our nice little selfie in front of a dog. And here we have. Also, this looks really phenomenal. Especially these are really love this picture. This seems a little bit messed up with the phone, but also this looks really cool. So I think it's phenomenal. So in this video, you have learned how you can train CCPD to make a prompt generation machine for me journey. You can either just copy and base the prompt that I showed you in the normal standard free CCPD interface, and CCPD will also write you some prompts. If you want to take this to the next level, use either the CPD that I have made or just train your own CPD on the data that I gave you. And if you are really, really smart, just train your own CPD and make the pictures perfect suited for you. Everything you have to do is to simply change the examples that I have included into my text file. So search for the perfect examples for your specific use case into the MurnyGallery, for example, and you can and will create a nice GPD. So go into the MurnyGallery. Here you can search for the pictures that you like the most. Give CCP, for example, six, seven, eight, nine, ten different examples of the things that you really like the most. For example, this one here or that or even this, if you want to make more a sketchy style or pencil drawing or this, if you really want to make something like an illustration, so you simply can train CCPD on all these different data and you will create perfect output, and by the way, I love this picture. This looks also relatively good. So my recommendation would be train CCPD to make prompts for you because this is awesome. You can create prompts in no time whatsoever. You have even included aspect ratios. And if you train CCP in the right data, he can also include stuff like Ks and much, much more. I tried these a few times, and I must admit it's may be nicer to include the parameters at the end of the prompt for yourself. But with the spectratio, CCPD does a pretty pretty damn good job, have fun. 63. Midjourney for Text generation inside your Ai Art: In this video, I want to show you that we can include text inside of difusion models. Yes, text generation with difusion models is a bit difficult. It works in Dali, so in Dali three with C CBD, but not in the platform. It works a little bit in Adobe Firefly, and right now in the version six of Murney it also works relatively good. Now I will show you two tricks that you need to understand if you want to create text inside of Murney. The first tip is, of course, you need to use the right style. You need to include dah dah raw because if you don't include dash R or the raw style, the outputs get not really that good. And the second tip is, you need to make the prompt really, really specific and not too long. Now I will show you a really nice prompt that includes everything that you need to create this. And then I will show you also what Mdurney under discord server tells us about generating text in this division model. So we started as always with the slash imagine command. And now you can see my prompt because his prompt works good. Close up of a woman with a natural face and hair holding a sheet of paper in front of her face. On the paper is written in handwriting the text I real. We even include a question mark. Let's just see if also the question mark works. The important stuff for me is that you need to make the prompt not too long, and you need also to include quotion marks for the stuff that you want to have inside the picture that should be text. So I real should be the stuff that is in our picture. Then the prompt goes a bit on natural and so on, hyperalistic, daylight and so on. The spec ratio is 16 by nine. This is important dash dash style raw, because if you don't use the raw style, the output is really not that great. Then dash dash stylized 250. This is a random number. You can even delete this number if you don't like it. And we use the version six. So important in this prompt structure is the prompt cannot be too long. You need to use quotion marks for the text. And the raw style works a lot better, and it only works right now in version six. And I would assume that in later versions in newer updates, this will get better and better. And now I send this prompt up. And while this is generating, I will tell you what Murne directly tells us about generating text with their revision model. And I hope you understand what we do to see this. So we press on this Murney bot in the left sideline here. And then we search on the announcements if we find something that is valuable for us. The announcements are always top up to date, and here you get always the newest information about stuff that is really cool. So you can see they have changed stuff like wired a little bit. The stylize and so on, will get change from time to time. You can simply read yourself what you have here because this changes like every other day. And here, we have basically a lot of information about the version six. And this right here is the text that I want to show you minor text drawing ability. So you can't create a lot of text. You need to use quotion marks. It should be the style raw, and then you can probably also lower the stylize values. So if you use a high stylize, it's maybe not perfect. For that reason, I used 250, and nothing crazy, so not like 900 and so on. You can even reduce the stylize to something like 50 and so on. So here, they have a nice little example, so imagine a photo of the text hello world written with a marker on a sticky note, spectral 16 by nine version six, and of course, you need to include the draw style, I think they have forgot it. So this is the important stuff right here. And now we go back into our small little server and see what we got. This is the server best di course. Of course, you already know that. And here we have our outputs, and I think nearly every single one of them are relatively good. The first one perfect. Second one complete mess. The third one looks also awesome and the fourth one is also really, really nice. I think the first one is the best. I want to make an upscale. I will show you this one last time. And now I also ask you if you see this picture like on the Internet, somewhere else, not in an AI course. Would you understand that is not real? Serious question. Think about it. 64. Permutation prompting, one Job Multiple Split Test: In this video, I will show you what premutation prompting is how to use it and why you should probably at least consider to try this out yourself. So the M Turney web page tells us about premutation prompting here the following. Everything we need to do is to include curly brackets and the stuff that we include in the curly brackets of this stuff, we will get separate chops. So if we take, for example, slash imagine, then a prompt, for example, red, green, yellow bird, we will get three separate chops, and the chops will be a red bird, a green bird, and a yellow bird. So the prompt is A, then curly bracket, red, green, yellow, curly bracket closed, bird, and then we get three different chops. We get one chop off a red bird, then a green bird and a yellow bird. I think this is relatively nice if you have a great prompt and want to run all of it in one simple chop. I myself, I use this to be honest only to compare different versions. I used this the following because I think the nicest thing to do is to simply use a fixed user prompt, and in the curly brackets, I include the versions. And with this cool little trick, we can compare versions really really nice. And even if we use the same prompt, you can see what's the difference between the versions. Because if you let a prompt run a few times, you will always get different outputs. To compare this really, really good, you need to use the same sat. Only if you use the same sat, you can see what really the difference between the different versions are. For that example, I will do something like this. We do slash imagine now a really simple prompt. A beautiful lake house in Alaska. Now I start my curly bracket, so we open it up. Then I start with version five, then am now we start with the next thing, dash version. Let's just say 5.2. Then one less time dash version six, da version and zero six. Now we can simply close our curly brackets, and I close everything with unspect ratio aspect ratio 16. By nine. And now I will also include a seat, d s and the sat is a random number, something like this. Now we can and will create amazing comparisons between the version five, the version 5.2, and the version six, with the same aspect ratio with the same sat with the same prompt in one single chop. So we send this prompt out and then M Cerny will ask me if I'm sure that I want to create three separate prompts. Are you sure you want to imagine three prompts from the template? This is the template, and then I can simply say added template, show prompts, no or yes. I myself, I always try to press show prompts to make sure that we did the right stuff. So we will run three different prompts. The first prompt is a beautiful lake house in Alaska, version five, aspecto 16 by nine and this sat. The second prompt is exactly the same, but the version 5.2, and the third prompt is everything the same but version six. So I think this should work, and I simply press yes. And I was a small idiot. So of course, dash d style doesn't work in version five. And for that reason, I used version 5.1 instant in this prompt. And now we can do our comparison. So this is version 5.2 in the first picture, version six in the second picture and version 5.1 in the third picture or in the third grid to be correct. And I think all of this looks really, really good. So also the version 5.2 is really, really stunning and really, really good. You see a lot of details, and I think this looks really nice. But if we go down on the version six, the version six is even more realistic, more crisp and we have a lot more details. You can also see on the water. The mirroring on the water or how you call this in English in German is called Dispgg This looks really nice on the water. If we scroll down, the version 5.1, yes, the version 5.1 is also relatively good. But the output is like not that crisp and not that good than in version six. And the stuff on the water, at least in this picture is not perfect. The details lack a little bit, and the version six is also not so good, at least in my mind. Let's do one last comparison with this same stuff. But of course, without the style raw, so we do something like let's just say a beautiful close up of a woman. Then I start with my curly bracket dash version 5.1 d version five dash version zero six. I close this bracket. I use the aspect ratio. I use the seed, but I delete the raw style because we can't use this. And now we will make a comparison between 5.1, five and version six, or you know what we do version four. So that you see how big the difference about four, 5.1 and six really is. We can also use like version four, five, and six that you see, how big of a chump the different versions are really. And here we have it. So this is version five in 16 by nine with the same seed. Here we have the version four, 16 by nine with the same seed. You also see the improvement from version four to version five. And yes, we can and probably should use better promises you already know this. I just want to show you what a difference this small tweaks make. So version five is a lot better than version four. This version four looks good. Yes. It's totally true, but we have this typical mid journey style they like to call it. And now we look at version six because version six is on another levelyf scroll down. Between my face, we have the things. So version six is enormous. Everything else is completely the same. Same prompt, same as batrato same seed, but the only difference is version four version versus version six, and the difference is enormous, and of course, if you compare this with version three, two or one, there is no comparison, and because of this, I don't show you version two, three, or one. You already seen what version one create creates for a mess because I already showed you this. You can't even see that we have a close up of a woman if you use version one. But my point of this video was the pre mutation prompting. Yes, it's a bit special. Maybe not a lot of people use it. But if you want to use it, just open up a curly bracket. Include a few different things. Just add a comma clute your curly brackets, and then rock with your parameters, something like the seed. If you want to compare versions or something like colors, if you want to make a lot of different prompts in one single job. And that's basically the pre mutation prompting. 65. Multi Prompting and Prompt Weights make you a Midjourney Pro: In this video, we talk about multi prompts. Multi prompts are really cool. You can do certain prompt weights or simply split your prompts. Let me show you quickly what Money tells us because they do a really good job. Multi prompt basics. Adding a double column, column column to a prompt indicates to the Mony bot that it should consider each part of the prompt individually. And here they have a nice example. You can simply use space ship as one prompt and you get the spaceship. If you use column column ship, you get a ship and some space. Here the next example, cheesecake painting. Here you have your normal prompt, then cheesecake, column column painting, Then you get a cheesecake and, of course, the cake painting. And you can also do this one more time, so cheesecake, column column, cake, column column painting, and you get this right here. And now I want to show you an even cooler example because I think this example was also included in the older versions of mid journey, and that was really, really cool. We start as always with the slash imagine. Now I do something like this. Hot dog. Then, of course, there is bat ratio 16 by nine. This is a simple prompt, so don't tell me this is too easy. Yes, it is. I just want to show you something. Now we do the same thing, slash imagine. Now I do hot I do column column I let a little bit of space and now I do dock, of course, also the aspect ratio will get included and we get really different outputs than previously. I think you are amazed by the difference. Here we have basically the two different results. First, the hot dog, I think everybody knows what the hot dog is, and this is according to Murney a hot dog. And now the next thing steps in because we can do a lot more with this. We can weight this prompt, so we can give the word hot, for example, more weighting than the word dog, and we can do it vice versa. So we have the prompt weights. And in simple terms, this means that we can include a number. If we don't include a number, the default setting will always be one. So if we scroll down, you can see here a normal prompt. So space column column ship is the same thing as space column column one and chip, space column column ship column column one, space column column two, ship and so on, so you can see if you include the same numbers, you have always a normal prompt. But you can make this different. If you use cheese column column two, cake. Cheese will get twice the weight then the cake. Of course, if you use cheese column column four, column column two. Now you have the same effect. So this should be the same prompt weight as this because four is twice as big as two, and also two is twice as big as one because one is the divult setting. And the same thing is true for this. Column column 100 versus cake column column 50. 100 is twice as big as 50. For that reason, it's the same thing as two versus the default settings of one. And that's basically the whole concept. And for that, I will go back into mid journey, and we will try to use a specific prompt weight on this simple prompt of this hot dog. So I copy this prompt. Then I do simply slash imagine. I include this, and now I will give hot twice as much weight as dog. And now all of this should get a bit hotter. I'm really excited to be honest to see what Meny makes out of this because you can interpret this however you want. So will there be fire it or is there simply more sun? Or is the dog simply hotter, what is hot dog? Now, I'm really excited what Merne gives us here. Maybe Mid ourney has the better examples with the space versus ship. Here you have space versus ship with one to one, so the ship is relatively big, and here you have space twice as much weight as ship, and you see the ship is nearly not there, so space has a lot more weight in this prompt. I knew it. Fire get added at least on this last prompt. So this dog seems to be really, really hot. And the other dogs, I think the prompt weight doesn't work perfectly with this specific example. Maybe Mid Cherney has changed this because space versus ship works better in this specific example. But for me, it's important that you understand how this column column works and that you also understand what prompt weights are. And you can wait not only your normal prompt. So not only you can split your prompt with this column column. After the splitting, you can weight in your prompt and after that, you can even wait your negative prompts. And the weighting of the negative prompts works something like this. Here you have your normal prompt, still life gauche painting. And then you have still life gauche painting, column column, fruit column column -0.5. What happens in this thing? You can simply see that the fruit got less weight in this prompt. So here you have a lot of fruits, and here we get 0.5, so we have the health of one, of the fruits. So roughly half of the foods get execluded from this simple prompt. I will show you this in journey directly one more time. Because, of course, we have also the dash parameter. I already told you how the dash dash no works. So you can simply type in dash as the normal parameter, and this is basically the same thing as if you would include -0.5. So vibrant tulip fields column column, Rd column column -0.5 is the same as red. But if you can be more precise if you weight this. Of course, you can do simply dash, but you can also do a little weighting of these. Now I want to show you an example. And here, you see some outputs. So our lake with our frog, da A R 16 nine. So this is simply a normal prompt with a frog. We have a lake with our frog. And then we have a lake with our frog, dash dash no green. You already know that we can do this. And this is basically the same stuff as we would do, a lake with our frog, column column green, column column, -0.5. You can also do just 0.5. You don't have to include this zero. I did this accidentally, but this makes no difference. You get really, really similar results. So this is the same prompt weight as if you would type in green. This is just a bit of the advanced prompting. But with this advanced prompting, you can be more specific. So you can increase this value. And I did this in the next prompt. So I increase the value to 0.9. And you see if we increase the value too big, the outputs will not be optimal. So something between 0.4 and 0.6 are values that work good for me. If you go up until 0.9, the outputs get a little bit ugly. Then we scroll down a little bit further because now I show you that you can also weight it like this. So our column column one with the frog column column two. So the frog gets more weighted on this prompt than we started previously. So you see the frog is more in focus than the lake. And you can do this vice versa. So you can do a lake column column two with a frog. You remember, if we don't include anything right here, the lake has twice as much value as everything, so the frog is nearly excluded from this prompt. And that's basically all that I wanted to show you. So one last overview because this is big. So you have learned that you can use multi prompting. You can include column columns. If you make normal prompts, so space ship, you get a spaceship. If you include space column column ship, you will make them different. You get space and a ship, hot and dog, cheese and cake. And much, much more. You can also do it like this as you have seen. You can do this however you like. After all of this, you can give all of this weight. So you can give the ship, for example, more weight, but you can also give the space more weight. The default setting is always one. If you use one, nothing will change. But if you use two, the thing will get twice as much weight as the stuff that comes after this. Four versus two is the same thing, and 100 versus 50 is the same thing because it's always the same mathematical equation. 100 is twice as strong as 50, just like two is twice as strong as one. And you can do this also with this negative prompting, so the normal negative prompting is. But if you want to be more specific, you can do it like this. Painting column, column fruit, column column -0.5, and you can exclude half of the roots of this picture. And if you want to exclude every single one of these fruits from the pictures, you can try to increase this value up to 0.6 0.7 maybe even 0.8, but 0.9 is most of the time a bit too strong. And this was basically multi prompting and weighting your prompts, and all of this is really important if you want to be really a pro in Murny. Just imagine you make some ice cream and you want to make your top with chocolate. Now you can simply increase or decrease the value of the chocolate with this column colum if you want to make this separate, you can also do it with this colum colum. Play with Codem codems, play with values, try to include different stuff. And in the next lecture, I will show you how you can do this also with pictures in your prompts. 66. Combine Multi Prompts and Prompt Weights with Pictures in your prompt: You already learned in one of the previous sections, how you can include pictures with the mi tool, but also into the prompts. And in the last lecture, you have learned that you can weight your prompts with column column. And in this video, I will show you how we can combine this. And this will be a quick one because this is easy because you understand all the fundamentals. This is always important for me. Step one, I upload a picture. It will be this picture of me, so I simply drag and drop it in right here. Then I can send this out. We need to have this in this prompting box. We need to have this here in this chord. The thing that I do right now is, of course, slash imagine. And now I drag and drop this down here into the prompting box. And now I press space. And if I type in my prompt, I can combine my picture with my prompt. You have already learned this. So I combine this with this nice little prompt. Linked in profile picture of a dog in a suit, professional photo. Of course, I can also not just type in dog, I can also include pull tarier or whatever you like. But I think this is fine for me. I don't use any Apexratio because the Apratio is one by one by the vault, and also this picture is one by one. So I simply sent this out. We work in version six with our raw style. So we send this prompt out right now. Now, this thing will get created. So this right here is our output, and I think the output looks hilarious. So this is me as a dog with a linked in profile back. It's some kind of a you see the style is nearly the same also weight. So everything we do for this is, of course, always type in slash imagine, then I include everything that we need. And after my link, I will type in column column, and now I can give this a weight. Remember, the default settings is always one by one. So if I type in two in this thing right here, My picture will get twice as much the weight as the stuff that comes after my picture. And while this is creating, I want to show you that I can also do it vice versa. So I include my prompt. I have simply copied this from above. Then I use column column. Now I use one right here, and then link in profile pic of a dog, and the dog should have twice as much the weight as my normal photo. Right now I will send this out. Right now, the dog should have a little bit more weight than me. So here you can see, I should have at least in theory more weight than the dog. I think Yes, it's a bit more human, especially the first one and also this one. To be honest, every single one of these maybe besides of these have a lot more humanity inside than the previous ones. So on the previous ones, you see, we have a lot more dog involved. Simply on this picture, we have a little bit of human. Every other single picture is like a lot more dog. And the second output is, of course, a lot more human like. You can see maybe, that's basically me. I think the ears are not that far off also my beard looks something like this, and the nose is also pretty damn good. I think this is like really human. This is just like me. And the last picture is also ready right now. And in this picture, we have done it vice versa. So my picture has only half amount of the weight compared to the dog. So the dog has the value of two, and I have the value of one. And you see also in the picture, you can see that the dog takes a lot of space into this picture. So we can simply wait also our prompts if we include picture in the prompts. In this video, you have learned that you can combine two of the coolest things inside of mid journey. You can, of course, upload your pictures and combine it with prompts, just like you have learned. Then you have also learned that you can weight your prompts. And in this lecture, you have learned that you can combine it. Upload a picture, throw it into the prompt box, use column column, and take a specific prompt weight if you want to play with this a little bit in more detail. This multi prompting opens up a whole word to try out into Murne. This makes you a Murney pro. 67. Save a lot of time with prefer settings, save your Midjourney cheat codes PROMO: This video is a nice little heck, because we can save and short cut a lot of our prompts. We can use preferred options, so we can simply save the stuff that we use a lot. Stuff that I at least use a lot are cameras and camera lenses. And you already know we can't rain chat GBD to make prompts, of course, and to tell us what lenses we should use. Let's just take one more look at this chat that we already had. We told him, of course, that he is a prop photographer and so on, and he tells us a lot about cameras, camera lenses, and much, much more. I tell him that he needs to describe everything as short as possible. And here we get, of course, a nice little camera for this specific task. And now I ask Chachi BD what camera he would use if he has to take your photo of a cat in a city. Let's just assume you are specialized in taking pictures of animals in different environments. Of course, if you are specialized about this, you already know what camera you have to use. If you are not specialized on that, we use CCPD for this answer. And CCPD tells us Asni A seven r, IV with F 50 millimeter lens, and F divided by 1.2 gm lens should work. I have absolutely no clue what this exactly is. But I know that we get good outputs if we include this into prompts. For that reason, I copied this, and now I paste it into CPD. Everything we have to do is to type in slash and now prefer here you have prefer out to them prefer option list if you already have a lot of stuff saved. Prefer option set, prefer remix prefer, suffix, prefer, and so on. The thing that you need is prefer options set. We need to click on these. Now you have prefer option set, and here you need to type in the stuff that you want to type in. And you see, I have already saved a view options here. I have a wallpaper, a camera and a co, and I want to call this, for example, camera or let's just say TCM. I think TCM is really, really nice. And now we need to press on this little plus sign right here, because now we can set the value, and we can simply paste in the value that I already copied from C CPT. Now we get this value. Now I press on the value, and in this value field here, I have to include the camera that I already copied inside of C CPT. So right now I have the Sony camera and if I sent this out into the chat, Mdurney will simply use this. We'll simply save this as our camera if we use TCM. So I sent this out, and now I have the TC and if I use TCM, I have this Sony Alpha lens and so on, and I don't need to include this every single time into the prompt. And of course, first, I have to remove the another because I already have a TCM. So I can also simply delete this stuff if I don't need it. Right now, our TCM is set and we can use it. And how we can use this this gets really easy. Now I make a small example. So we do slash imagine, for example. Let's just say colorful CT in New York. Then we type in CTC I think I have also spelling errors in the Cm. So of course, this can also mean something different, but we don't do this in this course. So this means TCM in my course. And if we simply use TCM, so we have simply made our own command right here. We can use this as a command and we can send this out. Of course, we can also include more commands. So the spectrti much more. Let's just do this a little bit better. And of course, first with DC Mall, thing that I want to show you. So colorful cat in New York. Now I start with d aspect ratio of 16 by nine. Then I use, let's just say, of course, the raw style version six. We have already included these. But I think a style of 750 would be nice and a little bit of chaos because I want to have different grids. So chaos may be of something like 60. This is our normal prompt. I will simply copy this prompt so that we can do a good comparison. So I sent this out right now and we will get a good result. I think at least the result should be decent, but the result gets a lot better if we simply include dash T C. Because then we include also our nice little lens and the result get a lot better, and all of this works really fast. So you can see now we have colorful T in New York, and then we have also our camera included with no Tsoever. And here we have our results, and I have to admit that also the first result is really, really awesome. So we didn't even have to include our camera lens to get really awesome results because I think this colorful cat is also beautiful. But if we use this camera, you get everything a bit more crisp. And because I use a lot of style and a lot of chaos, all of these pictures are completely different. So this cat is also awesome. I think every single cat of these looks awesome, and with this camera lens included this gets even better. Nobody can tell me that you can see a difference from a real cat in comparison to this cat, for example, or also that cat. They look exactly like from a prophy photographer. I think this is awesome. I simply love AI. I love Marney, please. You need also to love me and Marney AI because all of this is so awesome. Just make some pictures include shortcuts, do all of this. It would make me proud. Okay. 68. Include Smaileys in your Midjourney Prompts: This video is really, really a quick one. You can include smiles inside of your prompts, and everything works exactly the same as if you would include pictures. So you can do just simply slash imagine, and then you can include Smiles. You find the smiles if you go with the mouse over this thing right here, and then you can search for the smiles that you like. You can also type in Smiley, for example. And now I want to include for example, this Smiley, I send it out because we can make a prompt with this simple stuff. So one Smiley, nothing else. And remember, you can do the same stuff as if you would include pictures inside of your prompts. So you can do all of these. You can combine smiles with prompts. I don't want to show you everything just to save up some time. You can include smiles and use multi prompting. Column column one or column column two to wait your smiles. You can include not only one Smiley, but also two smileys. You can include a picture and Smiley, so you can combine everything else with this Smiles. And now we get this little prompt and why we get this output from this Smiley. Just look about this. This looks exactly like a picture from social media. You know that Marney is trained on a lot of pictures and is also trained on a lot of texts. I would assume that Marney is also trained on a lot of different pictures from, for example, Instagram. And on Instagram, they are always using such smiles and stuff. So if you use a Smiley, there is a high probability that you get an output of pictures that are also on social media. So think just about what hashtags or what smiles would people use on social media. And if you include such a Smiley inner prompt, you will create a similar picture. And like I said, you can combine the smiles with prompts, with text, with other pictures, with image weights, and much, much more, you can do a lot of stuff with the smiles. So I would assume that you simply type in slash imagine and use some smiles to have some fun. 69. Conclusions outlook and some Homework (copy my stuff): Over the last lectures, you have transformed into a Burney pro. You use all the right magic words. You have your own PD to create this amazing prompts. You can generate text. You use premutation prompting. You use prompting weights. You include Smiles. You basically saw everything. So you will do a really, really good job inside of Murne. Everything that we need to do next is, of course, to edit our pictures and to do our last fine tuning inside of Mere. And that's basically the outlook. That's what the next section is all about. But before we do all of these, I have a little bit of homework for you. Use one section, at least one and try this at home. I would suggest that you should copy all my magic words, and you should also copy my CPD, and then you are ready to rock everything in mid journey because you have a prompt generating machine, and then you should also maybe weight your prompts a little bit with this column column stuff. I think the possibilities with all of these are completely endless. 70. Finetuning and Editing of Midjourney AI Art: This section is about fine tuning your Murne AI work. We will do in painting or very region, as it's called in Murney. We will do out painting or as Marney likes to call it the Pan tool. And of course, we will also make some nice little face swaps if you want to have the best out of character consistency. You can even make AI influencers if you use the right facewapping techniques. We need to simply add the face swap bot inside of our Murney server, and we can search all of this on gTub. We can even zoom out of our pictures. And after everything is set and done, you can even share your whole M journey profile. So stay tuned for the next lectures because this is like the fine tuning of M journey, and then you are ready to share your artwork all over the place. 71. Midjourney Zoom Out to edit and expand your AI Art: In this video, I want to tell you that we have the possibility to zoom out of our pictures. Right now, as I am recording this course, you can do that only in version 5.1 and 5.2. But I am 100% sure that we will get this feature also in version six and also, of course, in later updates. So if you make an upscaling of a picture just like I did with this cat, you have only the possibility to make a sutile upscale, a creative upscale You can vary this picture subtile and you can vary this picture relatively strong. And now I will show you what we get if we do something else. I think we just use the same prompt because I think I like this prompt. So I copy all of this prompt. I include it right here. But now we will make some stuff a bit different. I think I let all of this completely the same, but only the version, I want to change the version to 5.2, and everything else is completely the same stuff. So I sent this prompt out right now. Now, this is the picture. I think I love all of this output, and I want to make an upscale of the first one. Because on this first picture, we don't have everything here from the ears, and I would like to have some stuff from the ears. So I press upscale of this picture, and then you will see the magic that we can do. So we can make, of course, all the upscalings and also very region about y region we talk in the next video. But we can do this right here. We can do zoom out. We can either press on the first button, and if we do that, we make this picture twice as bak. So I simply press on it, and then we zoom out and it will get expanded roughly at double the size. If you press this right here, you can zoom out and make it 1.5 time a big. Besides that, you have, of course, also the custom Zoom. And I think the custom Zoom is the coolest thing. If you press on custom Zoom, you can adjust all of this really, really good. So here you can see Zoom two. But you can do stuff like Zoom 1.2. So you zoom out something like 20%. And if you really like, you can also change the aspect ratio. And the last thing that you can do is, of course, to make this square. So you can simply press and make the square, and then you will also get a square image. And now we scroll down and see what we got. And here we have the picture, so the first one is, of course, the double Zoom. And you can see the picture got expanded by a lot. And of course, we have also a lot of variarty with this so moot because I have included this in the prompt with the style. You can see this picture right here in the left corner. At least to me, this looks somehow realistic, and the other stuff is a little bit more flowery. I would like to call it at least flowery. But I think also this looks amazing. And of course, you also see we have the ear right now. So this is the starting picture and you see right now we zoomed out twice. The next output is Zoom out 1.5 times. You see, we are a little bit closer, of course. So also this output looks really, really good to me. The reality in the pictures is enormous, and to the zoom out is completely seamless, so you cannot see that stuff got added to the pictures. Especially the ears are fascinating to me because on the first picture, of course, The ears were cut out, and in the newer pictures, you see the ears are always perfect. If I scroll down even further, you see this is my custom Zoom. Zoom 1.2. And you see we simply did a small amount of Zoom. It was just big enough to complete the ears, and the rest of the picture is nearly untouched. So this looks closest to the picture that we got, and we have some ears right now. And last but not least I can also make this square, and this is quick. We can simply press make square, and then you are ready to rock. This is basically the output from this square picture. So all of this look really cool. And every time the diffusion model gets used to include some new pixels. And of course, that's not all. Let's assume that you want to zoom out even further. You really love the stuff here with this flowers. You can make an upscaling of this forth picture. You can search your fourth picture, and then you can do the same stuff basically again. So zoom out twice zoom out 1.5 custom Zoom or x square. You can do whatever you like. I personally, I love the custom Zoom because we can simply type in the stuff that we exactly want to have. So you want to zoom out another 10%, for example, you simply type in 1.1, and then you have your custom Zoom. But for this example, I want to zoom out twice again because we want to zoom out a lot because I think we will get some nice flowers with these. And here is the output. So you see we have zoomed out really a lot and the pictures look phenomenal. I like I think every single one of these pictures, especially these right here. Let me know what you think is the most beautiful pictures out of these and do me a favor and use this tool. And in the next video, I will show you what the pan tool is because the pan tool is similar to this zoom out. 72. The pan tool to chance aspect ratios in midjourney: This will be a fast one because we use the pan tool, and the pan tool is really, really easy to understand. So you can see you have the things right here and what this means. If you press on this error, so for example, on this left one, you will simply zoom out, but you make only the zoom out in this corner. Let me show you. This is especially nice if you want to make, for example, a Tube thumb nail or something like this. So I sent this out, of course, and then we will get added stuff in this corner. And why is this nice? Let's just assume you want to make a thumb nail where you include text in one corner and where you want to have this stuff in the other corner a little bit. But of course, we need to consider. If you saw what I did, so I simply pressed on this thing right here. I also let the prompt here included, and this is one of the biggest errors that I see if people use this pan tool because you need to think always that this new stuff will get included. And now, let me just see if we have also a little kitten in this left corner, I scroll down. And yes, we have it. Okay, it's not entirely finished. But right now you see we have this. Of course, the picture is not 16 by nine anymore, so it means that we need to crop all of these. So if you really want to do this for a YouTube thumbnail, you need to simply make an upscaling, for example on this second one because I really like this, and then you need to download these and crop it with another software. Because the cropping in Murney doesn't work at all. So you need to go outside of M journey if you want to crop your images. But you see with one click, we can do this pan tool, and the pan tool works really, really nice. Of course, same thing if you press on this button. We will make your Zoom mot in this corner. And if you press on that button, we make your zoom or we add Bixels in this thing right here. And if you press down, we add bxels right here. Let me show you one last thing. If I press on these, of course, the prompting box will open up again. Sometimes this doesn't work. You simply need to press again because this has to work. And now we simply need to include the stuff that we want to see and delete the stuff that we don't want to see. We can delete every single thing if you don't include things. And everything you have to do is, of course, to add, for example, one simple word, for example, background. If we include background, we simply include background that we get added right here. So I include background. I should also include maybe like flowers. If I want to include more flowers, we can simply include the stuff that we really like that we think is nice if we use this pan tool. And the same thing is also true with the Zoom out. So if you press on Zoom out with one of these pictures you can also do by your custom Zoom, stuff with your prompt, so you can simply delete all your prompt and you can add other stuff. This is always the same key concept right here. And now I want to show you cool things. If we do, for example, custom Zoom, but also if we simply scroll down with this pan tool in this corner, we can add other things. And now I will show you that we can even add things that you wouldn't have. It would be possible. Let's just try dog. And here is our output. Yeah, so the dog was not perfect, but you see we have changed the aspect ratio. Right now, I think we have something like one by one as the aspect ratio. And the other thing is with the background, and this was relatively nice because we have included the flowers just like I would have thought. I think this output is also really, really nice, and that's basically the po tool. You can use the pound tool in mid journey to change your aspect ratio. You can make your outputs longer, and you can decide in what direction, the output should be longer. And you can also try to expand it up or down, and the aspect ratio will change every single time. If you want to have the exact aspect ratio again, you need to crop your pictures, but that you need to do outside of my journey. 73. Vary Region to edit your Image: Besides the zoom out and the pan tool, we can do really cool stuff with the very region here inside of mid journey. This is essentially the same thing as, of course, in painting and out painting in Adobe firefly and also in stable difusion. So everything we do right here is to simply press very region. By the way, I made here some dishes, some desserts, and I think they look awesome. By the way, this is version 5.2. Also, the version 5.2 can make really really enormous pictures. And everything we can do right here is to simply press very region, and now we can do the in painting here inside of this picture. This works completely different than the in painting in the other tools. You have either your squares thing right here and you can make square stuff over your pictures or you can use asso and make this a bit more round, you simply include everything that you want to exclude or to change in your pictures. Then you have simply to delete this prompt, and now I want to add just for example, background, and then this little thing will get completely deleted, and I sent this out. And here is our output and I have to admit the output is not completely perfect. So this get not completely deleted. You see, we have some stuff going on also right now. But of course, if you want to delete this completely, you can simply do this with another try. So you can simply do this one more time. And now I will show you that we can also include other stuff and not just exclude things. So we can press very region, for example, once again, Then we simply include things that you like. I will simply run with this one more time. So I include all of these, and then I want to try if I can do something like a handy. So we want to have a smartphone right here in this place. I try the word smartphone, and let's just see if M Courny can add here a little smartphone for us. And in the meantime, I let a new prompt run again because I want to show you a thing that is a bit cooler than all of these. So we use version 5.2. Of course, the aspect ratio, I will change it to 16 by nine. I have found this prompt simply in the MhurnyGallery, and I think the output got relatively good, even if we don't include camera lenses and nothing. Let's just see what we get from this output. Okay. And Murney messed up in this picture. So we don't have a smartphone. We have to be honest, excluded. Everything that was in the picture this time. Everything we have added in one picture is this beer. So I think Murney messed up a little bit with this in painting technique. But like I said, it's always important to me that I show you also the downsides of AI. Yes, of course. Not every single thing that you try will work for the first time. These tools are new, but these tools get better at a stunning speed. This will be always the worst variation, the worst quality that you will ever get out of the AI tools. Because they get better every single day. Now we have our picture, and I think I really like this picture, and now I want to show you what we can do with this. I make a small upscale off this second picture. Now I do very region once again, and I want to make he blonde, for example. For this thing, I simply do it with this tool. The lasso tool, I try to include every single piece of hair that I can, and I need to do this a few times, so you need to be a bit precise with these. It's always better to include a bit more than to include two little. So just make sure that you get everything what you want inside of your small frame and simply do it a few times. So we have here some hair and maybe also here a little bit. And now we simply delete every single thing from this prompt and simply type in blond hair, and I think this should work fine. Because this work, at least for me, nearly every single time I tried these. So we can make this blonde. Of course, we can also delete stuff from the background and so on. And yes, the painting in Adobe firefly, and also in Dali is easier to use, I think, and it's also a bit faster. So the interface for the painting in this court is not optimal, but I would assume as soon as we get our Mourny app, this works a lot better. And there we have it. Now she is blonde. And I think this worked really, really phenomenally good, especially this picture. So if I make an upscaling of the second picture, you can really see that the hair got nearly perfect. Maybe we have some brown hair right here. Maybe I forget to delete this brown hair, but of course, you can always try very region once again and delete the stuff that you don't like. Of course, if you want to include other stuff in this picture, maybe right here, you can always try these. You simply need to delete the stuff and then type in in your prompt what you want to see. In this video, you have learned that you can use very region inside of Murne to delete or change things inside of your Murney pictures. Yes, it is not that optimal to use, and, for example, Doby firefly, but it works, and I will assume that this gets better and better over time because this is the worst variation that you will ever use in every single AI tool that comes around your corner. This stuff gets better every single day. And as soon as we take this into the app of M Turney, I am sure that the in painting will work a lot better. Stay tuned because I will bring updates as soon as we have the app. Try this at home. I would really recommend it. Just edit your pictures a little bit. 74. face swap with the insightface bot in Midjourney: In this video, I want to show you some of the coolest things that you can do inside of M Journey if you want to edit your pictures. It's called phase swap, and we do this via this inside phase swap bot. Yes, we add another bot, so not only the Mourney but do our discord server, and he will help us to create really cool stuff. You can take, for example, your face and put it on Marney pictures, but you can also upload like pictures from you and make other faces over them. And of course, if you have an AI influencer, you can also use this tool to make this AI influencer really, really consistent. So let's just assume you want to have the maximum character consistency of a one single picture. You can totally do this with this bot. So, for example, you have 30 or 40 mid journey pictures, and then you simply swap the same face on every one of these and you have really consistent faces. For this tutorial, I will simply show you how we can make my face over another face over a mid journey picture. But you know you can do a lot of stuff if you want. So first of all, we want to make a small little prompt. I like to work in version six right now, and I do something like a man in New York. I want to do something simple. So we do something like this. Man standing in New York. Can and EOS close up brown hair 16 by nine, and the style is not war but raw. And we send this out. And in the meantime, while this thing is generating, I want to show you what we have right here. We go over to GTTub. GTAb is, of course, the programmer community, and we simply search for this inside phase. I will also link these. We simply scroll down until we get our link. So you can see you can make, for example, younger people older if you have something really cool in mind. You can simply swap new faces on other faces. But I think the coolest thing is to do, of course, to swap my face over a picture that we have created. We can do this, of course, with this swap ID. Things. But you will see everything. So we need to scroll down, and we simply need to press on this link, and then we need to include our inside face swap bot. So we simply press on these. Now, the bot asks me if it's okay for me if we add the bot to the best I course, and I think this is totally fine, so we press okay, terize and we are human. And here we are again. So right now I am in mid journey, and we have a new friend beside my face right here inside face swap. And now we simply need to add a character to our nice little bot, and then we can swap faces because I want to have my face on this guy right here. And I think this should work totally fine because we can simply make my face completely over this one, and then we are ready to rock. And how do we do this? This is really, really easy. The first thing we need to do is to save our ID. So we type in slash, and then of course, save ID. I simply press on this right here, and now I can simply upload a picture and I need to give this a name. Now, be cautious with this because you need to give this a name that is relatively short. Try something like me or he. I should have not more than five or six letters. So we try to make this easy, and I simply call this. And then we can and should upload the picture where we look directly in the camera, and I think I can see something right here. Now, yes, this is not a pretty picture, but we look in the camera, and I think this can probably work. As soon as you have done this, we need to send this in the chat. And done. Now we have saved our ID. So the ID is called, and I have to tell you something. There is the possibility that this doesn't work at the first try. And there are maybe two things that could happen if this doesn't work. First, the fort format. And for that, you simply go into a program that can convert pictures into PNGs online. This online converted.com tool is awesome. You can simply upload your pictures and then convert them into a PNG. And it's really that easy upload the stuff here and then download it as a PNG. As soon as you have your PNG, you get your picture back, something like this picture. And I hope you can see this picture is called. So I did a really short name. So don't include any Numbers and stuff like that because the inside phase pot has some problems if you include a lot of numbers. The two things you need to do. You need to make your picture into a PNG. If it's not a PNG, you go into a online convert tool. Second, you need to name your picture on your computer in the right way. Do something short. Then third, you need to make your name into this inside phase pot that is also really short and concise, for example, he or me or she or whatever you like. And the fourth piece is, if all of this doesn't work, just try to use another browser, because I had some questions from time to time that this doesn't work. So now we have created our ID, and I have to also tell you that you can create more IDs. I think you can create up until seven different IDs. So you can upload pictures from you and also from other people and make face swaps here inside of my journey. And if you want to make a face swap, you do something like this. Do you make an upscale off the picture where you want to make your face swap. For example, this picture right here. Now here, we have our small little picture, and now you need to set your ID. So you have saved your ID, and if you have more IDs than just one, you need to set your ID. So you type in slash set ID. If you press on these, here you type in the name. And of course, you need to see that you use the same stuff as here. And then you can simply send this out in your chat. Command send current ID name updated. So right now, our ID is now set with. So we have my face as the set ID. And like I said, if you have more than just one ID, make sure to use the right one for this swaps. And now you can press on the three dots. You go on apps and you use this in swaper and that's basically it. And this is our output, so I think my face got swapped relatively nice on this picture. Yes, this is not perfect, but I think this was okay. Now, let's test this out with the other two or three pictures. So we make an up scaling for example on this picture. So the third one. We have our upscaling right here. Then I press, of course, on the three dots. I use apps I use the swaper you see this works really, really fast, and it works nearly seamless. So I think also this picture is perfect. We have my face here on this little guy, and I think this got relatively nice. And now one last try, we try to make Superman, for example, those slash imagine Superman looking directly in the camera, and then we swap my face onto Superman. While this is creating, I can also tell you that we can swap two faces in one single picture. If you make, for example, a picture that has two faces, you can create, of course, two IDs, and then you can set two IDs. Make sure that if you set your Ds, the first one that you create, the first one that will get set will swap the face of the first person. The second one that is set will swap the phase of the second person. Now, this is a bit advanced. If you want to have auditorial about this, just hit me up, but I think this is not really necessarily like for a lot of people. But if you want to do advanced phase swapping techniques with this spot, just let me know, and I will make another video how to do this exactly. And here we have our superman, and I think I want to be the superman in the first picture. So of course, a small upscale. Three dots, apps in swaper in 1 second, my face will get swapped on this little superman. And I think also this output got relatively good. And with that output, I want to leave you alone. Because in this video, you have learned how to swap your face. Everything you need to do is to simply go onto the guitar page, press on the link add the insight swap butt, create your ID, upload a picture. Make sure you use a PNG. You can convert it if it's not a PNG. Use simple names. Then save your new ID. Press on the three dots of a picture. Go on apps. Use the insight face. That's basically it. And I hope this was really simple explained. So try this out at home because I think this is cool. And like I said, if you want to do this in an advanced mode, just let me know, and I will add a lecture where we swap faces of two or even three people because all of this is possible. 75. Share your amazing Ai Art: You have created a lot of mid journey pictures until now. And of course, in your gallery, you will find every single one of these. I think we started with this woman's right here. I think they look awesome, and then all the other stuff that we have created. And yes, not every picture is pretty that we made in this course, because, of course, I wanted to show you every single thing. But a lot of the pictures, they are enormous good. Here, as soon as we started to mess around with also the pictures inside of the prompts, I did not a really great job because I did use always simple prompting to explain to you how all of this works. But then we got also really nice results for example right here. And of course, everything you need to do next is to show people what you can create inside of mid Journey. Now, normally, you have a feature with that. You can share every single one of your artworks in no time whatsoever, and you just need to send one single link. I have to admit right now this is not working perfectly, so you have to try it with other ways. Normally, you have the button home on the left side. And if you press on home, you can simply go on this web page and send the link to people that you know. Like I said, this web page will get edited a little bit over the next weeks, and I hope we get this feature back because then you are ready to share your AI work with one single link. And until then you can simply go on social media and post some of your AI works. Just make an upload on Instagram because some of this picture can really go viral. If you make really, really cool and good content, I am sure people are more than willing to see what you have created. So have fun creating this stuff, and I would really recommend you to edit and post the stuff that you have made in my journey. In that manner, have fun posting your stuff, and I am excited to get the share link back and as soon as we have the share link, I will make a small update. 76. Conclusions and Summary of Midjourney (Plus how to get free fast hours): Over the last sections, you have learned every single thing that you need inside of M journey. And in this video, I will make just a quick, quick summary of all of this so that you have everything once again in your brain. I think we can learn better if we do this. So we take here this M Journey guide to make this quick overview because they list everything out relatively nice. Of course, we started as quickly as possible, and we make a mid journey profile, we selected a plan, and of course, we also created everything inside of this chord and added our bot do our own server to make our creations. The quick note of the community guidelines is just work with a bit of brain and don't be a jerk. We also dived in into prompts. You learned that you can add text, but also pictures text and parameters to your prompts, and it's always important to include this stuff in your prompts. M Journey gives a lot of different examples for prompts. We have the possibility to make variations out of our pictures if we press on the right buttons. With the very region, we can simply edit our mid journey pictures, and we can combine very region with the remix mode just like you have shown. We have made brown hair into blond hair. We have a lot of up scalers. Every single one of these works different and not every single one is really good. And we have different versions. It starts by one. It goes up until 5.2, and also, of course, the version six is awesome. That's right now the newest one. We work in this court and you have learned everything that you need to know about this court. You should also join their community and up to date all the time. You have learned every single command. This is the whole command list and you know everything. If not, go here on Md journey. The important stuff is blend to describe user info settings and present show chop shorten, and of course, the most important stuff is imagine settings and maybe also blend. You have learned all parameters. And here is the whole parameter list with nice descriptions. You know the aspect ratio chaos, no quality, repeat, seats, stop, style, stylize, tile version, video, and wired. We have more tools. We already talked about very regions up scalar and styles, but you also can use the pan tool and zoom out of your pictures. You have also learned the advanced stuff. So multi prompts image prompts re prompt weights and much, much more. You know that you can use the describe tool, the blend tool. Here is Mourn little bit repetitive because we have done all of this also at the beginning. You have even learned the premutation prompts, This right here, the remaster. We don't use this. I think you don't need it. You can see for yourself, if this is something for you. It works right now only in one version and you can simply make new redos. If you'd like to call it that way, nobody uses this feature. And also the stop. Not a lot of people use this. So you can stop the generation at any time if you want to have like bad pictures. You can add dash dash stop ten or dash dash, stop 90, and the diffusion model will stop before the picture is finished. You have also learned everything about the subscription one small dip for right now, you can earn 3 hours if you go to rate images. So if you make your image ratings just like you have learned, you can get some free credits. The top 2000 here of each day will get one free hour. And last but not least here son policies. And I myself, I am not a lawyer. I just want to tell you don't be a jerk, make cool stuff, and don't try to harm people. And that's basically all. And the conclusion of all of this inside of Murney is Murney is an enormous AI tool, and my recommendation is use it. So please use Murney to create amazing AI art. So your homework is start generating pictures inside of M journey. When you start, you are not that good, but you will get better if you do more. So create some pictures inside of M journey, and you will be a pro in no time. And of course, this course will be the fast track to be a pro. Because you know every single thing that is possible inside of M Journey, and you can always look back at the lectures if you get stuck at a time. And I have once more tip for you. If you could leave the course of five star review, that would mean the word to me. If you have learned something, leave it just this five stars because other people will see it, they will also come in the course, and together we can learn more. Maybe somebody will come into this course just because you leave an ice rating, and maybe this person is the missing link in your business or in your career with Mourney. Together, we will make better images. I know it. And if not, just have fun creatings on pictures with M Journey. 77. The difference between Stable Diffusion, Midjourney, Firefly and Dall E: The next sections will be about stable division. And in this video, I want to talk about the differences between stable division and all the other division models. So you know, stable division is a division model. But what's the difference between Dali, Adobe fire and Marney? The biggest difference is that stable division is completely open source. You have already learned that we can use the code of stable division, and stable division has the so called checkpoints or if you want to call it just simply variations. And we can train these checkpoints for ourself. We can download checkpoints, and we can use all of this in a lot of different interfaces and web applications. This is the biggest difference. We can even fine tune our own Laura, and you already learned at the beginning, what our Laura is. Just go back into the basics if you want to get a bit deeper. But of course, we will make this also step by step over the next lectures. Here are some key insights. What's the difference between stable division and all the other division models. Like I said, it's completely open source, and there are a lot of ways how we can use stable difusion. One of the things that I like the most is Leonardo AI or even perhaps C art because we have really cool interfaces with Leonardo I and Cart and this makes it a lot easier to use stable difusion. Because of that, I want to start with Leonardo AI. In Leonardo AI, you can start completely for free, and if you want to have access to every single feature, yes, you need a small subscription. But like I said, most of it is free. Same thing is, of course, true in C RT. Also here, you can do most of the stuff completely for free, and if you want to do certain things, you maybe need a subscription. Then we have stuff like Automatic 11 11 and CF UY. You need to install this locally. So you need a really, really strong CPU, and for that reason, at least right now, I will not include automatic 11 11 and CFUY in this course, because I think most of the people can't use this because their PC is not strong enough. But we have also the possibility to use Google Clap. You know what Google Cap is, and we can run the code the open source code of stable diffusion inside of Google Cop, and that is also enormous nice because we can do this all completely for free. Okay Google CLP will give us the GPU power that is necessary, and the code is also free because stable division is open source. So over the next lectures, you will find a lot of stuff. Most of it will be completely for free. You will always find a free alternative if you use stable division, even if you don't use automatic 11 11 or CVI. Now if somebody really needs to have automatic 11 11 or CFI, I can also make updates in this course how to use this two applications. And I will totally do that because I get the newPC with a really, really strong GPU. So for me, it's no problem to make updates about this stuff in the future. And the coolest part of all of this is we have a lot of Google Colab notebooks for video generation. This is enormous. We have stuff like Dvorum or rep fusion, and we will make really enormous animation with stable division. So stay tuned for the next lectures and the next sections because all of this is enormous. Stable division is a really, really cool AI tool. It's one of the best at least in my mind, because we can do really everything that you can imagine. We can train our own models, and we will also do that inside of Leonardo AI and also with Google Colab. We can create pictures. We can create videos. We can do every single thing that you can imagine, we can do in painting out painting and much, much more. But we will start really slow. We will start with Leonardo AI. As soon as you get every single thing inside of Leonardo AI, you will get what stable difusion is, how it works, and then we can make a deep dive and also train our own small AI models because we can train our own difusion models. So stay tuned for the next lectures because this is awesome. 78. Reminder, Stable Diffusion Checkpoints, Models and LoRas: Real quick. How stable division is different than the other AI models. I just want to show you how the model is structured. So how stable division is structured. We already talked about this a bit in more detail at the beginning of the difusion models, but this is a while back. And because of this, I want to tell you once again, how stable division is structured. We have one gigantic division model. And of course, you have something like variations. So you have, for example, just like in mid journey, the variation, one, two, three, and so on, but in stable division, they are called a little bit different. We have stable division 1.5, for example, we have stable division el and so on. These are simply different variations of this division model. And some of them make good realistic pictures, other make good pictures, for example, for animus and so on. And these variations are called checkpoints. So every single checkpoint make different variations. We have the Chan out checkpoint and much, much more. And you can't only use the checkpoints, but you can also use auras. A Laura is a small thing here inside of such a checkpoint. And the Laura is something like a parameter inside of mid journey. So you have a small small aura, and this Laura is, for example, trained, especially on animus, for example, or on faces or on spiders. So this Laura is trained on a specific thing. We have, for example, stable revision Excel. So if you want to call it the variation or better terms the checkpoint, and then we have a Laura. And if you combine a really good checkpoint with a specific Laura, you get a really good output. And the coolest part of all of this is we can train our own checkpoints and auras with stable fusion. That means we can inject ourself into the AI. We can inject pictures of me, for example, inside Allaura and make pictures of me. I already showed you one or two pictures that I have created with Allaura from me. But we can also train Aura on animes on Pokemon on whatever you want to create. And that's the coolest thing about this stable difusion stuff. So we can train our own models, and of course, we can also use models from other people. And the easiest way is to use Leonardo AI to start with all of these. And because of this, the next section will be about Leonardo AI, and we will do a deep dive. So stay tuned because stable difusion is awesome. 79. how to get access and the interface of leonardo AI: Welcome in this section of Leonardo AI. Leonardo AI is a really cool tool. It's an interface. It's an app if you want to call it that way that works with stable division. Leonardo gives us simply the interface for stable division, and you can use it at least at the beginning completely for free. You get some credits every single day to test all of this out. If you want to have the better subscription, of course, you can make this. Now I will show you how you can get access to Leonardo AI. Everything you has to do is to simply Google Leonardo AI and press on the first link or use the stuff that I give you. If you don't have an account, you will be on this webpage, and you simply press, create an account. And then you simply need to type in your e mail address, and you will get an e mail to confirm your identity. It's really that easy. As soon as everything here is done, then you are in an interface that will look something like this, and I will give you a quick overview. So this right here is the interface, and the interface is really, really easy to use. You can see in the left corner there is Leonardo AI, and after that, you see your credits. Right now, I have 8,500 credits because I have a subscription for Leonardo AI. But if you start out, you will get 150 credits every single day, and maybe in the first day you get even a little bit more. You have home, you have the community feed, personal feed, training and data sets, fine tuning image generation real time Canvas. Real time gen, Canvas editor through texture, and so on, and you can even get access to their API and much, much more. So you can simply press on every single thing that you want to explore. And if you want to make this quick, you go with the mouse over the stuff that everybody uses all the time. So you can generate pictures. You can use the real time Canvas. You can use the Canva editor, and you can use motion to generate videos. If you press on image generation, you can create every single thing that you want. So you make art, illustrations, landscapes, and also, of course, photos, and so on. With the real time canvas, you can simply draw what you like. This is also as soon as I will show you this. With the editors, you can do simply in painting and out painting just like in the other AI tools, and with motion, you can create your own videos. Now, if you scroll down a little bit, you have here a nice gallery, just like in Mhurny and you see the outputs of Leonardo are amazing and really stunning. Especially stuff like pictures and creative things are awesome. But also the photo realismus is really, really impressive. You can simply search for yourself for stuff that you really like. And as always, you can always press on every single picture. You can see the prompt, You can also see one stuff that is different in stable division than in the other division models. You can also see the negative prompt because in stable division, we use negative prompts nearly all the time. And you can also see stuff like the resolution when it was created. Some of the things that are included. So in this example, alchemy, don't worry, we will get over all of these. You can see the sat and all the settings. And of course, you can simply create with this model if you press on this. The model is the alberdo basXL model. Like I said, there are a lot of different models inside of stable division. You can also see the stuff right here. So image to motion, image to image and remix. If you press on remix, you can create the same picture as this user has created. So you will be in this interface, and every single thing is included that also this user has included. If you simply press generate, you will be able to generate nearly the same picture as the user that did this previously, of course. And there we have it, and I think this looks awesome, and this is the easiest way to create a stunning picture inside of Leonardo AI. You can, of course, always go back. Don't be overwhelmed by this interface. This is really, really easier to use than you think. If we go back, of course, we are again in this gallery, and you can always search for the stuff that is interesting for you. So you can sort by trending by new and by top. You can search for all motion, photography, animals, architecture, character, food, and sci fi. So if you press on motion, for example, you will get the coolest videos. And if you press on the videos, you will see how they look. So you can simply search for the coolest stuff and see how all of this looks. Also, this is a video that I really like. You can simply search for the stuff that is most interesting to you. All of these videos are really, really cool, and now all of this video start to get in motion for us. And that's basically the easiest start of the interface. So if you want to simply start to create your first picture, just go on trending stuff that you like, for example, you go on all, and then see for yourself what you got. You can simply press on one of these pictures and create the same thing. And now over the next lectures, we will do every single thing step by step. So how to write prompt how to do all of this in the right order so that also you can generate the perfect pictures for you and don't have to just copy stuff from the gallery. So stay tuned for the next lectures. 80. the pricing of leonardo Ai, start for free: Let's talk about the subscription inside of Leonardo AI. This will be aqui qua. Everything you have to do is to simply press on your credits in the left corner. And then you are on this web page, and you can see what plan you actually have right now. As soon as you start out, you are in the free plan. And in the free plan, you get 150 fast tokens every single day, so you can create some pictures completely for free every single day. Right now, I have this subscription where I pay 12 bucks a month, and I have 8,500 fast tokens every single month. So that day but month. I can create up to five jobs at the time. I have private generations. I can train ten models. This is the coolest stuff for me. I can retrain them, and I have priority infrastructure. Then you have also the better plans right now, I don't think that this is really needed. It goes up until 60,000 fast tokens every single month. Unlimited image generation at the relaxed mode, unlimited video generations at the relaxed mode, limited up scales, up to 20 jobs at a time, private generations, we can train 50 models per month, retrain 50 models per month, we have priority infrastructure, and we have up until free concurrent generations. That's basically all about the plan. You can simply search for the plan that is right for you. If you use pay yearly, you get a little bit of a bonus then to pay it monthly, and that's basically it. And if you ever want to change something, just press on managed subscription, and then you are ready to go. And that's basically the subscription. I would assume that you should start out in the free plan. And if you ever need a paid plan, just use the cheapest one because we can do nearly every single thing with the cheapest plan. And if you want to do stuff also completely for free inside stable division, we can also use stuff like Google Colab and much much more. But the interface inside of Leonardo AI is really awesome. And because of this, I simply use the subscription because I love the tool. 81. The first Images with leonardo AI and Stable Diffusion: Now let's start with the image generation inside of Leonardo AI, and of course, we use stable division, and you already know that we need to use the right model inside of stable division. So the easiest thing to do is to start out with the right model. So yes, you can press on image generation, and then you are instant inside of this interface and you can create your images. Here you see the models. If you press on these, you can search for the right model. But I think it's easier if we do this the other way around. So we go back one more time. And then you press on this right here, fine tuned models. Because here you have the platform models, community models, euro models, and favorite models. Now, euro models right now, you don't have nothing because we haven't trained the model yet. Favorite models, most of the time you have also nothing right now, but we have platform models and community models. And you can simply press on them and see what model is right for you. I would assume that you should start with the blood form models. Because here we have a lot of different models, and you can simply see for yourself what you want to create. So we have, for example, Onardo version Excel, Leonardo difusion Excel, photoo stable difusion 0.9 and much much more. And everything you have to do is to simply press on a model that is perfect for you. For example, let's just say you want to make photo realistic pictures. You simply press on the photorealisms. And then you get the examples, what you can create with this model. You see you can create really, really cool pictures with this model. And everything you have to do is then to press generate with this model. You can simply search a model that is perfect for you before you use it. And now we come to this small interface. So we start in the left corner. Of course, this is how many credits you have. Then you can also say how many images you want to make bare generation. I just say it start with two because you don't need eight pictures if you just start out. Then you have photo real. So you can include this or you can also exclude this. It depends on what you want to make. If you include this, your pictures will get really photo realistic even with a simple prompt. Let's just do man and we press generate, and by the way, this is, of course, the prompting box. So right now we have included photo realismus with our model, so the Leonardo photo real, and we will create a horrible prompt. The prompt is simply man. Of course, you don't have to use photoalisms. You can include alchemy. And if you don't know what the stuff does, you just go with the mouse over the question mark. And the question mark tells us right now, Leonardo Alchemy is our powerful new image generation pipeline. Click the question mark for more info. And here you can read every single thing for yourself. In simple terms, we get better output and stronger contrast with these. You can include and exclude every single thing that you want to do. And of course, with one mouse click, you can do this. If we exclude alchemy, we can also use prompt magic. This is basically the same thing. We get a little bit of better outputs if we include this. So you have high contrast and the prompting strength. And if you don't like it, just exclude it. Then you can make your images public or private. This is if you have a subscription. Then you can also use the image dimension. For example, now I want to have the image dimension in 1024 by 1024. This is one by one, but you can also use stuff like 16 by nine. And all of this and you can adjust all of this also right here if you just make these things bigger or smaller. Then we have this guidance scale. If you go with the mouse over this, you can see how strong your prompt is weighted. I personally I use guidance scale between five and maybe nine. If you go to high with these, the outputs will be really awful. And also Leonardo tells you what you can use right here. Every time you get this symbol right here, something is not right. You go back to maybe eight and you see this thing gets excluded. The same thing is true by the image dimension. Every model is always trained on different image dimensions. So if you use, for example, an image dimension that is not perfect for the model that you are using, Leonardo will tell you, Hey, maybe it's better to use a different resolution. And of course, if you don't want to do this, you can also rock with the resolution that you want. Right now, Leonardo tells us 1204 by 764 would be the right move to do. So we need to use this resolution for optimal output with this model. Then we scroll down a bit because we have also the possibility of tying just like mourny and we can simply include the tiling, and then we get the output that we want. If I simply type in Leo part, we get here our tying stuff, and we can always use different models. Of course, here are models that I have trained myself and so on, but we will get to this. Right now, I will make a Leo part stuff in 16 by nine because we can do this. Then we have the advanced settings and also image inputs and so on, but we will talk later about all of these because we can also include pictures and much, much more inside of our prompts. On the advanced settings you see, recognize we can exclude this if you don't want that Leonardo reminds you of the stuff that is not perfect, for example. But I like to include this because, like it's really practical. We can also use a fixed seat just like in Mourney. So if you use this, you can use a fixed seat, and then you are ready to create character consistency. If you exclude it a random sat, we'll get used. And you can also use a little bit of different scheduler. And if you go with the mouse over this, this defines how the noise affects the generation. I personally, I love Leonardo. But you can also use oiler. I don't love all the other stuff. So Leonardo or oiler is the stuff that I like the most. I would assume you just use Leonardo, and then you are perfectly fine. And just like I said, you can always use here the model that is perfect for you. If you already used a model before you generated your first picture, this is perfect, and if not, you can simply use your model right here. Let's just use the dream sharper for example. And then you can include the stuff that you like. I include right now, for example, the alchemy, and the rest, I think is nice in this settings. Then we can say what we want to create. Beautiful men on the beach. We will make anime with these. Then you can also go on this right here. So you can use dynamic, anime and so on. So for this example, I simply roll with dynamic. We can add elements more about the elements later. This are simply the auras, and we can also add a negative prompt. And about the negative prompt, I will talk in the next video. Because right now we are simply generating our first pictures, and I think you have seen now all of the standard stuff inside of Leonardo AI. And here we have our output from the dream sharper version five. Beautiful man on the beach was the prompt. And you see, we have not really exactly what we want. And part of this is because we don't use a negative prompt. Negative prompting is really, really important in stable division. If we don't use negative prompts, the output will not be that great. And because of that, I want to talk about the negative prompting in the next video. But in this video, you have learned how to start with Leonardo AI. The important stuff is that you use the right model. So use the right model and then use the settings that you think are necessary. All the settings are somewhere right here on this side, and you can simply go with the mouse over every single setting and let Leonardo AI explain exactly what they do. And of course, over the next lectures, we see exactly what's important to generate really good pictures with stable difusion. And like I said, the advanced settings like sat and so on. All of this works just like in mid journey. But the biggest difference is the negative prompting. And about the negative prompting, I want to talk about in the next video because without the negative prompting, we will create ugly results just like we did right now. So we made three pictures and none of these three pictures are good because we didn't use negative prompting. So make sure to see the next video because this is really important in stable division. 82. Prompt Engineering and negative Prompts for Stable Diffusion: Let's talk about the prompt engineering inside of stable division, because the prompt engineering and stable difusion is a little bit different than in all the other difusion models. The first thing remains always the same. You need to be specific in your prompt. So your normal prompt is as always. You need to include the subject, the environment, the lightning, the colors, the background, and so on. So you include the normal stuff as you always know in your normal prompt. Nothing changes here. But the big difference is the negative prompt, and of course, also, that you need to search the right model for you and you know how to do this. You press either on image generation and search your model inside your image generation interface, or you press simply on fine tuned models. Now you search for the right model for you. Let's just use this model. You can see what pictures we can make. I really love a picture like this, so I think we should generate with this model. The base model is here stable division. This is simply a bit fine tuned. We press generate with this model. Now we are again in this interface, and now I will show you a cool trick. Let's just assume that you want to create this beautiful colorful woman. So it was something like a painting of a colorful woman. And you don't know how you can make your prompt better, even if you do know, but I want to assume that you have forgot every single thing that I told you about prompt engineering. Now, Leonardo AI is here to help you because you can press on this prompt generation. If you press on prompt generation, Leonardo will help you create better prompts. You can simply copy your prompt and then enter your basic prompt idea right here and press here on Date, and you will get four ideas that will create relatively good outputs. So a vibrant woman with a ketoscope of colors oring her body. Her hair of canvas, of swearing, hers, and so on. I think this English is a bit too much for me. But I think we should generate this prompt because I hope this works like fine. If we press on this, we will get a prompt and everything else will get included that we enabled right here. Right now, we don't enable alchemy. So I think we should enable this because this makes better pictures normally. The pictures are public. This is okay for me. The image dimension is also okay for me. Guidance scale. We can use it maybe at seven, eight is also fine, so we can just leave it. But I don't want to have dying. We had included tying and dying is awful for this. Then we scroll down recommendation size. So all of the rest is fine, I think. Now we press again and we don't get tying stuff for our output. And the output will be relatively good, I think. But the thing that lacks right now is, of course, the negative prompt. And there we have it. I think the output is now really, really good, especially these one. I think they are really cool. And now let's just see how we can make this even better. Of course, you can play with all the other settings. But for me, it's important that you understand how to craft the prompts for stable vision. We need to add negative prompts. And right now you can also see that a really, really big negative prompt is here enabled. This prompt is from a previous generation from me. Here we have enabled too much beard. Big beard, destroyed eyes, destroyed hands, destroyed mouth, de de but, twisted hands, twisted legs. Blur and so on. So we have a lot of stuff included that we don't want to see. And I have made a nice list for you so that you know what you can include. You include basically all things that you don't want to see, so you can create a picture and then look at the picture and see what is not that good. And most of the time, if you create bigger pictures like from whole people, it are hands and eyes and such details. I have made the best things here into one simple list, and then you can simply include, to be honest, just every single word of these, and you will get a lot better outputs. Plastic, deformed, mutation, muted, extra limbs, poorly drawn face, poorly drawn hands, missing limbs disfigured, cross dies, blurry, floating limbs, and so on, and I have always a small description. So you can simply scroll down onto this PDF, and then you can simply include all these words that make sense for you. Out of focused and so on, all of this is really nice. Watermark, writing, text, noisy grain, and so on. You can simply copy some of these words and then include them in your prompt and you will get a lot better output if you include this in your negative prompt. Even if you just include stuff like extra limps, poorly drawn face, and so on, the output will get better 99% of the time. Right now, I press generate because I think this will make the output better, even if we just include standard words. You don't have to be that specific with your negative prompting because it's just important that we add some words like out of frame, long neck, poorly drawn fingers, and so on, so that the diffusion model understands what we don't want to have in our picture. And there we have it, and don't tell me that this output is not a lot better than the first one, and all of it just because of the negative prompt and the negative prompt was made really, really easy. It was not specific. I just copied the stuff from my PDF right here, and if you want to be specific, you can also do this, of course. For example, just look on the picture and think about what you don't like in this picture and include it into your negative prompts. But this standard words, they are always perfect. So muted extra limbs poorly drawn face. Missing limbs disfigured, cross dies, blurry, floating limps, and so on, this is always stuff that you can and probably should include in your pictures. And that's the concept of negative prompts. It's really that easy. You just need to include negative prompts and then type in the stuff that you don't want to see. You can be specific. So for example, if you make a human and the human has a long neck, include long neck in your negative prompt and the neck should get shorter. Most of the time, you can just include the stuff from my PDF. Poorly drawn hands and so on. This is the standard stuff that works really, really well. And with that, you will get better outputs. I can nearly guarantee that because stable division works better with negative prompts. So in this video, you have learned how to write prompts for stable division. The normal prompt work like in every single division model out there, and Leonardo can give you more ideas with one simple click. You've already seen that. Besides that, you need to use negative prompts. Stuff like ugly, blurry outer frame, poorly drawn hands, bad eyes, bad anatomy long neck, and so on, work all the time. Even if it's not specific for your picture, if you just throw in these words, your output will get better. It's just how it works. 83. Leonardo Help Guide: And one more trick. The help guide from Leonardo AI is completely awesome. Just look with me at this. If you press on CU and help, you are on this web page, and this web page is really, really cool. Here you can press on home page navigation, getting started with the AI image generation tool, prompting tips and tricks, and also FIU help guide future releases, and, of course, also the API, because we can call Leonardo AI also via API. But I think the API is too expensive, and that's why we don't do a deep dive into the Leonardo AI API. So you can simply press on all of this and see for yourself if there is something valuable for you. If you press, for example, on help guides right here, you get eight different articles how to use different stuff. If you press on alchemy, you will get the best practices out of firsthand, out from the Leonardo AI team. They have also some videos in slow motion, how you can use this and much, much more, and they simply tell you some stuff that is important for all of this. Now, I can't go over every single thing of this because the course will get out of the hand quickly. We can talk about this for days and days and days. But I think you learn everything that you need in this course, and if you need some more info, just go on this web page. Like normally, you don't need it, but you know that you can do it if you really want to dive even deeper. Like I said, normally, I simply show every single thing that is important for you. I just cut the fluff. 84. testing prompts and using the Zoom out, remove Background and upscaling: In this video, we will create some really good pictures, and after the picture creation, I will show you what we can do after we have made our pictures. Because we can also do up scales and much, much more inside of Leonardo. I want to start with this prompt. This will be a young woman. I add a really good negative prompt. It's basically the same negative prompt that we used right here in this thing. We used a fine tuned model, so the Leonardo diffusion Excel, it should be dynamic. We have, like I said, the negative prompt, we use alchemy, and this right here is the resolution, and everything else is at the default settings. I press generate. The important things are that we use a relatively good prompt, and of course, that we use a good negative prompt. Then we play with alchemy and stuff like that, and we will get a good picture. This is the output, and I think the output got really, really good. Let me show you some more prompts and settings how you can generate good pictures. Always remember, use a good model. So it's really key that you use a good model right here. You need to play with these, and if you don't know what output they make, just simply search right here. You go back on Leonardo and you go on fine tuned models and search the stuff that you love. And right now, we have a lot of cool pictures. Just take a look with me at what we have. I tried some prompts for myself. I tried to copy some prompts. I tried to remix some prompts. I tried to use random prompts, negative prompts, some stuff, of course, also from the prompt generation right here, and we got a lot of really, really cool pictures. For example, this right here. This is simply the illustration of this cat. Then this is also really nice. And remember, we use always different models. So this was absolute reality, for example, but we used the style right here of illustration, so we can always play with all of these things. Of course, if you use, for example, anime or creative or dynamic, you get different results. If you use, for example, me right here and you press generate, you will get a completely different output. I hope you understand this. So this is an illustration. Then right here, I used also another model. This time, it was the three D animation style with the anime model, and a really simple prompt. You see, we can create really cool pictures, and this was completely different. It was the Aldebo base XL model with dynamic included and a positive and also negative prompt. Here I included some brackets, for example, long neck and elongate neck. Because at the first try I get a really long neck. And if you want to make some things stronger, you can include brackets in your prompts. This right here is a nice little sticker, and this is also really cool, so this stuff works amazing well. The Phoenix looks also nice and here the picture that looks really realistic. If we scroll up, you see right here the absolute reality with the anime looks just simply a bit different than the illustration. And what can you do if you think one of these pictures is really, really nice. You can simply press on them. And if you press on them, you have a lot of options. If you press on these, you can download this image. If you go on the second thing, you can copy these to your clipboard. So if you press right here, this will simply get copied to your clipboard and you can include it into other stuff. Now, this is the Zoom. Right now, this is in beta. And if we press on this, we will zoom this image a little bit. Let's just try this out. And now the Zoom is done and you can go into the left corner and see for yourself what we got. So with this right here is, of course, the original image. And if we press on the unzoomed image, It looks something like this. So the zooming was really, really nice. We can go back on the original image, and we can simply see that the zooming is really, really cool. It is just similar than in mid journey. Besides that, we can remove the background. So if you press on this button, we will remove the background of this nice little illustration. And of course, also here. You can switch between the images. So the original image and also the image with no background. If you press on this, you can see, we have the picture right now with no background. Even if this thing is right now loading, and it seems that this is not finished right now. You can simply press on no background. Normally, this is finished, like, really, really fast. Besides that, you have also this things here. You can use alchemy smooth up scalar or the alchemy crisp up scalar. Of course, it makes more sense to use this on the original image. If you go back on the original image, you can do up scalings here. You can always see for what this specific up scalar is. The alchemy smooth up scalar is for images with smooth textures, which include illustrative anime and cartoon like images. This will cost five tokens. And the crisp scalar, this is more for texture details for photos, digital art, and some free D renders. Also, this costs you five tokens. And this right here is new. We have the alchemy up scalar. Use this to refine and upscale your images. This can improve face and hands as part of the process. So if your picture is a little bit messed up in the face or even the hands, this up scalar can probably help you a little bit. And if you really hate your picture, you can simply delete it by pressing on this small button. So in this video, you have learned that we can first of all, create amazing pictures. If we just play a little bit with the prompts, with the negative prompts, with different models and with different settings that I already showed you. So I would really recommend you to play a bit with all of these settings. And as soon as you have something that you like, you can do a lot of stuff. You can make three different variations of upscaling. You can zoom all of this, you can remove the background, and of course you can delete your stuff. So have fun trying this out. And in the next video, I will show you how we can include Laura's to bring our picture generation at a new level, so stay tuned. 85. LoRa in Leonardo AI strich Elements: In this video, I want to talk about Laura's stable division or in Leonardo AI. In Leonardo AI, they are called elements, but it's basically the same thing. I hope that you know what a Laura is. So we can use a checkpoint or a model, however you want to call this. For example, Leonardo model and so on, and then we include a Laura. And this Laura will fine tune our checkpoint. I will show you right now how we can do this. And for this example, we do it really, really simple. So we use this prompt right here, so photorealistic woman and so on, and we don't use any negative prompting. We don't use stuff like photo alsmus and nothing. I simply want to generate this picture with this simple prompt in the Leonardo vision Excel model. The model itself makes really, really good outputs, and because of that, we don't include other stuff. And we also don't include other stuff because I want that Duras elements have more weighting on our stuff. And this is the output. The output is relatively good. Yes, we can make this output a lot better by including stuff like photo realismus or even negative prompting and so on. But I want to show you what the elements can do, and we see this better if we don't include a lot of other stuff. So we add an element right now. We press plus on this sign. And here you see what elements you can use with this thing. You have right now five different auras that we can include. And here you also see that this is called Laura right now. And you can include four Las at a time. Let's just include, for example, this here, the coloring book. If we include this Laura, we will get the style that looks something like this. I press confirm right now. And now you see we have this element included. This Laura is included. If I press generate, we will get a style that looks something like this, and we can increase and decrease the strength, so the weight of this aura. Always look at this stuff right here. A weight of zero will mean this element has no impact on generations. So if you take zero, of course, you have no impact. You can use negative values and we create something with the negative values. If we go to zero, we create a picture just like with no weight. We already created something with one, and now I will increase the laura to the maximum amount of two. And here you see the difference. So this is the normal output. And here we have included our Laura. Here we have really completely different outputs. This was with the Laura settings at one. So we get a really strong impact if we use the settings at one. If I scroll up, here we have the Laura with negative laura weightings. And you see this negative laura weightings, I have to admit they don't work that good. I personally like to use Laura weightings that are between 0.8 and 1.2. Everything else will yield results that are a bit I would like to call it different. But you can play with all of these, of course. The picture with the Laura weight of two is still creating. And the cooled part is right here. You can press on elements every single time and you can include more than just one Laura, of course. The vintage Christmas illustration, Laura, in combination with kids illustration. I think we will get the cool output if we include this two Laura. So right now we have two auras, and we can always play with the strength of the Laura, so I confirm this two right now. The strength right now is at one of both of the auras, and I press generate with the same prompt. Sometimes if we include Laura's, the output can take a long time until it gets generated. And while these pictures are generating, I will also show you that you can go back in Leonardo, and all of this will still generate. So we can go for example, into the showcase, and we can search for a picture that we like. Let's just say we love this right here. We can remix this and our stuff right here will still be generated. And now we could make a picture out of this. And if we use another model, let's just say, for example, the dream sharper version five, we have other lauras that we can include. So if we press on elements in the dream sharper, You see, we have a lot of other different models. We have the crystalline Laura, we have the inferno aura, and we have much, much more, and we can always select up until four different auras. This is important because every single model that you use have different auras. Let's just use this crystalline Laura, for example, and create one more picture. I press confirm right now, So we are using the stream sharper, and the prompt is a bit different, and I will include this crystalline Laura. I press generate with the Laura weight of one, and then of course, one more time generate with our Laura weight of zero. So with an excluded Laura. And here we have some of the outputs. And remember, I always want to tell you the truth. So the truth is that sometimes this stuff fails, at least right now. Remember this is always the worst stuff that you will use forever. All of this will get better. So we have some failed generations, but the other generations were fine. We have this Christmas Laura and it worked really good. Here you can see this crystalline aura. Here is the crystalline Laura increased to one, and you can see the crystalline is crystal. If you go up right here where we decrease the crystalline, you see that we have excluded it completely. I have decreased the weight to zero, of course, and now we have a normal picture without this laura. Remember, you can always include more or less auras. Just go on the stuff that you like. You can use also Biobank Celtic bunk and for example, the green stuff right here, and you can see selecting more than two elements may have unexpected outcomes. So yes, you can include up until two four. But of course, if you use crystalline in combination with toxic bunk and then also biopunk you know that this maybe is not the best solution. So in this video, you have learned how you can include as inside of Leonardo AI. They are called elements. You can simply press on the button and then include up until to four different auras to your prompt. Remember, more than two elements will most likely make the output the bit special to call it that way and you can increase and decrease the strength of this aura. If you put it at zero, it will get excluded. One is the default setting, that is relatively strong, and I think settings between 0.8 and 1.2 do the best job because everything else makes the output not like the prompt. Play a little bit with different auras, combine different luras and you will get enormous good outputs inside of Leonardo AI. That's basically all that you need to know about auras. 86. Image2image: You can upload pictures into stable difusion and then include them in your prompts. The stuff that we need to do is called image to image. So we upload the image to create new images, and we can influence the strength, so how strong should the image be recreated in stable difusion? Now, you can upload pictures of yourself or also pictures just like that have a good style. You can upload every single picture that you like, and we can regenerate them. Let's just take a quick example from a picture of me, and you can also search for other pictures that you have some examples so you just want to make some good prompts. I do a really easy example. I search for a picture of a man, but we can also search for a zombie or whatever you like. I think this right here should work, and also that should work. Let's just start with these and maybe after that we make something a little bit scarier. So I press on these So we have the right model, we have the right negative prompt to the right positive prompt, we use alchemy and also the right resolution. Now, if you scroll down right here, image inputs have moved to the image guidance tap, and the image guidance tap is right here. And if you press on these, you can see that you can upload four different pictures. Now, let's start with uploading one, and you can also select from recent images. So if you press on these, You can simply see the pictures that you have used. I, for example, I used a picture of Arnold Schwarzenegger and also of Sam Altman. And here we have some generations. But we can, of course, also just drag and drop new pictures in here. I just want to upload this picture of me. Now the picture is uploaded, and now we use this step right here. So image to image. And, of course, in the next lectures, we will also take at depth to image edge to image post image and text image input. Right now, we use simply image to image because this is easy to use, and everyone can use this. You see, right here, we have a little bit of a problem. So select aspectrratio, doesn't match the dimensions of the image. Match your output aspect ratio for best results. So this picture right here has 512 512 pixels. And for that reason, we need to use this also in our output in the left corner, so I go down onto 512. By 512, and then the aspect ratio should be perfect. By the way, you can also increase your aspect ratio. You can also do, for example, 1024 by 1024 for the output. So something like this, it's just important that we need to make a square picture. Now we have here something like this. In the description, we can, of course, do everything that we want. So direct the color patterns and the overall entire look view of an input image, and we'll use this to guide your image generations. Here we have the strength. So how strong should my picture influence the picture that we make? If I increase this to example 0.9 and I press generate, we get nearly the same picture as I have right now. So I press generate. We can, of course, go down with this strength. So if I go down, for example, at 0.7, my initial picture will get edited a little bit stronger. So I press generate again at the strength of 0.7, and you have always to go to the generation history to see what happens. So you can simply switch into this thing right here and then see what you create. So the first picture, this is 90% me and the prompt has really, really little influence on my output. But the second picture here, we have a little bit more influence on my picture with the prompt because I decreased this strength a tiny bit. Right now, it's also 0.7, so my input is really, really strong right here. If I go down to example for 0.4, you will see that the prompt will get executed stronger and stronger. If I go down to 0.25. The output will get even more like the prompt. And of course, the lower I go, the less input my image has on this prompt, so we go down to 0.10 right now. And here you can see it. So the prompt gets stronger and stronger. Right here, the prompt is even stronger and this right here is nearly just the prompt. So my picture has really, really little influence on this thing right here. And of course, you can use other images. You don't have to use pictures of you. I just wanted to show you how nice this works. So 90%, 70%, Roughly 50%, 20% and 10%. You can do this really, really easy, and you can simply go on this image guidance and include other pictures if you like. So we can just select another random image. I press select from recent images. Now, I press show More. I can either upload images See my generation go in the community feet or use follower feeds. I go simply into community feet for right now, but all of this is really nice also to create logos. I also made some Tube logos and mid journey logos and much, much more with this tool. You can do really a lot of stuff. Right now, I want to go in the community feet. In the community defeat, you can simply search for trending new top or whatever you like. And by the way, here, you find some really nice pictures. So this bird looks awesome to me. The snake is cool. All of these pictures are really, really cool, at least in my mind, and you can simply search for the stuff that is perfect for your pictures. And I think I need to take this one because I love this picture. So right now, we use this picture. We press confirm, And now we have included this picture right here, and you can see the following here. Image to Image cannot be combined with other image guidance for XL models, photoreal and PMV three. So another checkpoint. But right now, I want to exclude this because I need to use this picture because I think we make cool stuff. We use image to image once again. Then we need to take the right aspect ratio. So we have one by two, and we can simply use this. So we use one pi two so that we have the right aspect ratio. Everything is included. And now I want to run with a really, really simple prompt so that you see that all of this works easy. I simply want to make something like B two. Pokemon. I hope some of you know Bccu. It's a nice Pokemon, and we can simply play with different styles right here. We will use this style to make bicchu or bcco or however you pronounce that in English, and I simply make a few generations here with this image input because I think this should work really cool. Now, let's go onto the generation history and see what we got. Here we have our outputs. This right here should be 20% image weight. So you see we really have Pikachu, the Pokemon, but the style got used from this picture. If I scroll up, this right here is 70 or 80% the original picture, and you can see we have nearly just the picture. This right here is 90% and you don't see nothing from Pikachu. If I scroll up once again, this is just 10% of our input image, and you see we have only Pikachu left. You need to play a little bit. I think something like 30% would be awesome for our stuff right here. Maybe also a bit higher like 37% or something like that, because I want to have the style, so the style is really nice as you can see right here. A picture that is a really nice combination is at least in my mind, here, the gold standard. So you simply want to use the style from your pictures and then get something really cool that is completely new. So this is way too strong. You don't have to do stuff like this. But this picture is enormous cool because we have our prompt, but the style of the picture will get included. Now, let's just see what we get out of Pikachu right here. Also, this looks phenomenal. This was now like something like 30%. Maybe you know what I used. I think it was like roughly 30%. And if I scroll up one more time here, we get 37% strength, and let's just see what we got out of these. 37% seems at least to me a bit too high because the face of B ktuG changed too much. All the style is perfect, so maybe something 30-37 is perfect. Something like this looks really, really cool, and also the first picture is enormous. So in this video, you have learned that we can use image to image. We can simply upload an image, and we can simply mix it up with prompts, and we can choose the strength of the image. If you go up until 90%, we have just the picture that will get used. You can upload pictures from you, pictures from the gallery, and to be honest, every single picture that you can find wherever you want, and you can completely regenerate them. The lower you choose your value, the higher the prompt weight will be. And at the end, if you only go to 10%, nearly just the prompt will get executed. Most of the time, something between 20 and maybe even up to 40 or 45% is really cool if you want to include the style. And if you want to make stuff that is really, really similar, then your picture, just go with something like 60 to 80%. If you use ten or 90, you just have the prompt or just have the picture. So don't use this stuff unless you want to because there are no rules in stable diffusion. 87. Leonardo Ai, Controlnet, Depth to Image in Stable Diffusion: In stable diffusion, we have a really, really cool feature that is called control net. With the control net, we can simply control the outputs. So we can use the depth of a specific output, the edge of a specific output or the pose of a specific output. And I think you understand what all of this means. We can simply upload a picture and use this specific thing for our generations and combine it with our prompts. And of course, we can also use the different strength. And in this video, I will show you what depth to image does exactly. I have generated some pictures, and I think the pictures are really, really cool. This right here is a button, so the Tube button. Here you can see the picture that I have uploaded. And I simply tried to play a little bit with the prompts. I have increased the strength to add the maximum capacity. The prompt was simple, so colorful explosion. The negative prompt is out of frame aglibrafm This things got this nice little output, and I think it's awesome. And here we have exactly the same thing. So colorful explosion, but the prompt was winkles naked nudity, nipples, two heads, extra fingers, muted tends, poorly drawn hands, and so on. So you see the negative prompt is not that important. It's just important that you include some words that you don't want to have in general in your outputs. So it don't has to be that specific. I would argue that stuff like mute tens poorly drawn hands and so on, got us even better than the specific stuff like ato frame aglibrary and deformed. So you need to simply try a bit with the negative prompts, and you can simply include the stuff that I already told you. We also played a little bit with this model. So right here, this is the Leonardo division model X L. I used this for a few examples and always the prompt colorful explosion. Here I did no negative prompt at all. I think the prompt with the big negative prompt in this model got us the best results. Now I will scroll down to show you that the dream sharper model did not deliver that great results. So you need to play a bit with these. Leonardo Division XL did a really great job, and then I changed the prompt. Right now, we have the prompt phase of a lion. And in this picture, the strength was a bit too high, so we don't see nearly nothing about this lion. Same thing here, but then I reduced the strength even further, and now we get a nice picture, and this was the first picture, but now we can't really see the YouTube button anymore. So you simply need to play a bit with all of these, and you do that by simply pressing on image guidance, and then you need, of course, to use depth to image. So you use depth to image. This is the control net feature, and then you can search for pictures that you want to take. So you can press on these and you can also change these pictures. You can upload different stuff. You know how this works. If I use this one more time, I can simply search for new stuff if I want. Let's just go onto the community feet and search for one more picture where we want to include the depth. I really like the depth of this picture, and because of that, I use it. We press confirm. And then, of course, you play with the strength. A strength of one is not that strong. And then you also need to consider always the aspectratio. So we use the spectratio two by three to get the best output. And then we don't use colorful explosion. Yeah, we can use this. Let's just try to image colorful explosion or like colorful explosion bird And that's basically it out of frame, and so on the Leonardo difusion model. Now be press generate, and I want to make a few picture of these. I increase the strength of a tiny bit on every single output, and we should get the depth of this picture on our new prompt. And of course, you can also search for better prompts and use this stuff. So you can simply go onto Leonardo AI, You can search for the output that you like, for example, this. I want to press remix right now. So we have everything that got us to this picture. Now I use the image guidance. I need to make sure, of course, that we use once again, the depth to image with my input image, and we press confirm. So we use the prompt and we should get the depth of this image with the prompt that we have. I increase the depth a bit and now let's just see what we have generated right now. So, of course, first, we started with the depth to image, colorful explosion and the bird, and here's some negative prompts. But we get the depth of the image that we had in our control net. Then of course, the same thing. But the control net was stronger, and you can see we have really exactly the depth from our image and the same thing basically right here. And now let's just see what the new prompt get does. So here we have the silhouette of artic queen and so on. So here, the depth is not perfectly included because the depth was just the strength of one. Of course. If you use a bird here in the prompt, it's really, really easy to use this depth. If we use a completely different prompt, it gets harder and harder for the division model to include the depth into this prompt. And if you increase the depth to the maximum capacity, just like in this prompt, you can at least see a little bit of this bird. To be honest, it's a bit of a mess. So yes, you can increase the depth or the strength to two. So the maximum capacity, and we simply get a little bit of this depth from this image. So it works really a lot better if you use stuff also in the prompt, what you have in your depth to image field. Let's just see what we get out of the prompt if we increase the depth to the maximum capacity. So the prompt here is completely different. So silhouette of an tic quin filled with a forest. Double exposure, crisp lines and monochrome background. Of course, it's not easy to include the depth of a bird right here. You see the bird here in this field, and it's not that easy to include this. And right now, the division model is also really, really slow to create these. This is the maximum depth that we can include if we use a completely different prompt, and I think this is, yes, it's creative, but I don't like the outputs. So make sure to use a prompt that is near to the input image that you use, and you can just use depth to image and everything that you like. So just use depth to image if you want to have the depth of an image that you found that you think is really cool. Increase or decrease the values, and that's the depth to image with the control net in stable division. And in the next video, I will show you the other features that you can use with this control net. In Leonardo, the names are sometimes a little bit different, but control net can do these things, and depth to image is one of the nicest features of control net. Besides the post to image. But like I said, in the next video, you will see the other features of this control net. 88. Controlnet, Pose to Image and Edge to image in Stable Diffusion Leonardo Ai: In this video, I want to talk about the control net one more time. We talk about Edge to image but especially post to image because post to image is really awesome. So if you look at this right here, we have depth to image, I already showed you this. We have image to image. This is not the control net. We have g to image. Edge to image is relatively nice, and you will see this awesome in this video, and post to image that's the stuff that I want to focus about in this video because post to image is really awesome. Text image input, I will show you this in the next video. This is not really the control net. Let's talk about post do image, and we do it via this way. Now, let's assume that you want to make a really cool picture, and you simply want, for example, to create something creative. Let's just say this one right here or also that one. Let's just see what other people like so that we can create something that other people really love. We go on to the top. I think we should generate something like this. For that reason, we simply press remix, and then we can copy all of these. Now, if I simply press generate with this, of course, we will regenerate nearly exactly the same picture, and I want to increase this to two or three so that we can see more of these. Press generate one more time. So here we have the output. Here are our three pictures. I think the pictures are really, really cool, but now we want to add a pose to these. Maybe the neck is a little bit long, so we can, for example, include in the negative prompt add the first thing like right here, long neck once again because I really think this is a bit of a problem of this prompt. And remember, if you want to make this stronger, you can use this right here. So we can simply include some brackets, two or three brackets, and this will have the strongest weight. So long neck for our negative prompt, and then we do really cool stuff. We go on image guidance. Now we use image input, of course, and we simply search for something. We go to see more. I like to go onto the community feet. And here we search for a post that we want to have this woman in. Let's just try it with these posts. I think this would really look awesome. So we try to use this pose. And here, we can use the strength, and also the aspect ratio must be in this format. So three by four is the aspect ratio. And now we need to make sure that we don't use image to image, but we want to use pose to image. So we need to include the pose in our picture. The default setting is one. So I press generate with the default settings. Then I can increase it a little bit to 1.3. I press generate once again. And I can also increase it even further to 1.9 and I press generate one last time. We should be able to include exactly this pose. Yes, this is maybe not the best pose for our prompt, but I think this should look at least interesting. Let's just see what we got. And here we get our outputs. So you see, we use the pose exactly like in the picture. It was exactly this pose, and of course, it depends a bit on the strength that we use. I just let it run a few times. Some outputs are not that good and other outputs are a bit better. We really fixed the long neck, but some of this stuff is not that good. Now, maybe you ask why this is not that good. And I will tell you exactly because you need to use the same style in your prompt then in your post to image. This is how the control networks. I also showed you this in the last video with this bird. So right now, I want to show you what we need to do exactly to get the best outputs. First, I want to search a photo realistic prompt. We press on photography. I use something like this. Now we remix this prompt. Now we search for a post that we want to have this prompt in. So for example, a little bit of another post and we got in our picture, but the style should always be nearly the same. Maybe something like this would work really good. So I had to try to run with these. We press confirm. Now we need to use post to image. We have already used this, and we need to use the right aspectrto once again. So four by seven. And now we press generate and see what we get. Of course, we can increase and decrease this strength once again. I increase it a bit, and then I want to increase it one more time to the maximum strength so that we should get at least the post from this image with our new prompt. And here we go. The first one, at least in my mind is the best one. We have exactly the pose from our image. So we used the new pose but exactly with the old prompt. We have Mgfox somewhat like a robot exactly in the pose from our pose to image control net. The second one was not perfect. Now, maybe you ask, why is this not perfect if we increase the control net? This is a bit tricky sometimes. We need to play with this. The last picture is with the maximum amount of the control net. And here we have also the pose relatively good. So you see the original pose is this right here, but I would argue the strength of one got us the closest results. And now let's just see what edge to image is because edge to image is really, really similar. Stable difusion just tried to search for the right edges in your picture. So let's just see what we can do with these. We press, of course, once again on this image guide. Then we switch from pose to image to edge to image and let the diffusion model just simply guess what edges we have in this picture and we press generate. Of course, we can increase and decrease this. Also, in this selection, the value of one so the default settings work most of the time really, really well. I increase it to 1.27 or 1.3 roughly because in my experience, also 1.3 sees the right edges with these. And then we see what we got at the generation history. And this is the first output, and I would really argue that the diffusion model got the edges right. I think we get really the edges out of our picture and meshed it up with our prompt. I think this picture looks awesome. And if we scroll up, I have increased the edge a little bit to 1.3 roughly, and also here we see that we get the edges really, really nice. I would argue these two pictures got really the edges from our input picture. And of course, go just on these and play a little bit with these values and also a little bit with your prompts, and you will get awesome results. So you have seen both depth to images, Edge to image post to image. And in the next video, I want to talk about. Text to image. Because in this video, you have learned that the post to image simply changes the pose of your character. You need to be cautious because the character in your input image with your pose should be similar than the stuff that you want to create. If you want to create a full body view of a woman in another pose, you need to make the prompts and also the input image somehow consistent. Just the pose will change. So if you use completely different stuff, you will also mesh this up a lot. And with Edge to image, basically the same thing, stuff that dissimilar and the division model will search for the edges and include that in your pictures. And in the next video, we will see what we can do with text. 89. ControlNet Text Image Input in leonardo Ai is it broken: This will be a quick one. Let's talk about text to image. Everything you have to do is to go as always onto the image guidance and then upload text that you want to create. We can simply drag and drop our text here in. So this is a picture on the picture is simply the text text. It's important that it should be on a wide background. The aspect ratio is seven p two. This is not optimal for Leonardo AI. Let's just see if we have something that is at least a bit similar. I think this should work. And now we can simply include in our prompt, what we want to see. Fire and ice is something that I think should work and the strength of 0.3 at the default setting, so we simply press generate and see what we got because this works really, really nice most of the time. I increase it a little bit, press generate again, increase it a little bit and one last time press generate. By the way, this was simply a screenshot from word. So I simply opened up a word document. I typed in text and made a screenshot out of these. And now let's just see what we get. Okay, I have broken Leonardo AI. Normally, this should work. Let's just see what I got out of this. I tried to play a lot with the prompts and with a lot of different stuff. I also deleted, nearly 20 of these, and I just get out all the time this thing. So normally, This should look something like this, text image input. I hope you see the difference. So you upload a picture of black text on white background, and then you can simply make your prompts and you get stylized text back. Right now, it seems that Leonardo AI has excluded this feature from their tools. To be honest, I have no clue why, let me know if this works as soon as you use this because I am 100% sure, they will bring this feature back because I used this feature to be honest a lot and this feature is really, really cool. So let me know if you can generate better text than I did in this tutorial, and of course, I can make an update as soon as this works again. Normally, I had no problem, but right now it seems that this is not included. And like I told you, all the time. I want to also show you the downsides. Not every single thing works all the time. This should work, but right now it seems that it's obviously not working. Normally, we upload these, we play a bit with the values and we type in in the prompt the stuff that we want to see. I played with colorful fire and ice, fire and much, much more, but none of this seems to work right now. I hope you have more luck than I have. 90. Character Consistency thanks to the Seed, making PEOPLE older & a bit of Prompting Fun: In this video, I will give you just a view of the tricks that I learned in Leonardo. Yes, we can include seeds. You already know what a seed is. And now I will show you how we can use the seed and also how I use the alchemy. So you see right now, I already have created a picture, and I will always blend myself out of the frame because I think it doesn't look nice because this interface is a bit full with me. So you see, I have a cool prompt. I used the photo realisms. And with the photoralism, we make really cool pictures. Of course, I use here the cinematic view. You can also use creative vibrant or none. We use no element. So no Laura for this. Now comes the cool stuff. Everything that you can and should do is simply to play. So you can simply include the stuff that you like, and if you don't like it, just take it out. So right now, I don't want to have photoalisms and I press generate. But we work with alchemy. Then I don't take alchemy anymore, and I press generate once more time. Then I press. This right here makes no difference. This is just if we can see it or not. And then we can always use different styles. So Leonardo style or none. And then I include the hooalismus back in. We can also include photo realismus and if we have included this, the alchemy will get set automatically. But instead of cinematic, we use vibrant, for example, and make one more picture. And then I come back to cinematic because now I show you the seed. But first, let's just see what we got over here. So first, we started, of course, with the phooalismus, and we used simply a normal prompt without negative prompting and nothing. If you use phooalismus, we can't use any negative prompting. Then we did it without the photo realismus but we get also a nice picture, at least in my mind. So you can simply play a bit. If we don't use alchemy, the prompts and the outputs are really a lot worse, at least in my mind. So play with realismus with alchemy and without alchemy to get the most out of your pictures. And now I will also show you that we can do cool stuff if we go on to the advanced settings. If I press on these, I can simply use a seat. So let's just say the sat of 888. I press generate, and then I press generate once again, and you know what happens if we do that, of course, we get the same picture. And if we use the same sat and the same prompt and I press generate again, we should be able to get the same picture twice, exactly the same picture. And right now, also the other stuff is created. So this is the picture with the hotter alsmus but with brand instead of cinematic, also, this looks nice. And here, you see, we have exactly the same picture. So if we use the same prompt, Of course, all the other settings the same. And then we also use a fixed seat. We get exactly the same picture twice. And if you want to have character consistency in different environments, just change the environment right now. So not perhaps in a sunlight beach, just like say in in New York. City. Yes, the rest is okay. So now we should get a consistent character, but, of course, in New York City and not on the beach anymore. And there we have it, so you see, we have the street view behind her and not the beach anymore, but the rest of the picture is nearly exactly the same. I think this is completely awesome. It worked like magic. First, two pictures, same seat, same prompt, and then same sat, and I changed the prompt just a tiny bit. And now she is not on the beach anymore, but in a city with people in the background with a street. And here we have some sand and other stuff, and I think this is enormous cool. So in this video, you have learned two things. First, play a little bit with alchemy and different settings. It makes fun. You see relatively fast what create the best outputs. Second, use a fixed seat if you want to have maximum character consistency. And of course, you can also make her older. Let's just see this one more time. So right now she is in New York. This is completely fine. But right now she is 35. I press generate, then I make her 45 55, and the last thing, 65 or even 75, if you like. And there we go. Right now, she's 25. Here, she is 35. And you see, it's exactly the same picture, but she looks really a little bit older. Then we scroll up. But because right now she is 45. Yes, she looks awesome, but she is a little bit older. Then 55. So you can see the minor changes. This is really, really awesome. She looks maybe not really like 55. Yes, Leonardo makes a good job to make her look young. Then 50 65 years old. And right now she really looks a little bit older and 75 years old, and I think my goal is to look like her as soon as I am 65 years old. Just look at her. She looks awesome at 75 years old, also at 65 years old. Yes, you can argue this looks a bit too young. Just make her like 95 years old, if you think she should really be a little bit older, but you see how well this works. You have maximum character consistency. Every single thing is completely the same. She just gets a little bit older with every single prompt. It was never so easy to get such character consistency in any diffusion model than in this right now. Come on, 95 years old. This is not 95. My goal is to look like her as soon as I am 95 years old. 91. Tiling in Leonardo: Tying in Leonardo, I. This will be really, really quick. Step one, use a prompt that you like. For example, fruit lat. Use an aspect ratio. Go down. Include tying. So this stuff right here. Press on generate, you get something like this. You press download on download image. Then you go into the seamless texture check like Mn, you already learned these. We upload the picture to see if it works and you see it works. It's maybe not completely seamless. Yes, it's relatively nice, so you see it really worked. Here you can see a little bit of a mess, and if it doesn't work, you just go back and you try it one more time. If it doesn't work, try again. I think my failure was the aspect ratio. If we use one by one, we get better outputs. Right now, I make some roses. I did basically exactly the same thing, and now I press download on this roses and the roses seem to work. And that's the tiling in Leonardo, AI. I have to admit the tiling in Murne is a little bit better. Just play a little bit if you want to, or you use Murney for the tying. At least I do that. 92. Summary of the Stable DIffusion basics, outlook and a bit homework: This video will be a quick summary of what you have learned and also a small outlook into the next section. You have learned how to get access and what's the pricing model. So you can start for free. Just go on their web page. We can create our first pictures with really simple prompts. If we want to make this better, we need to use the right model and some negative prompting. We can and probably should also include stuff like alchemy, photo realisms, and much, much more. After the generation, we can do up scales, download the pictures, make a small zoom out and use all this upscaling stuff. Leonardo and stable division has a lot of different auras. Here in this section, we simply used pre trained auras, and of course, we will also train our own auras later. Then we took a look at image to image and also at all of the aspects of the control net. Control net is really awesome. We have depth to image H to image. Post to image, and we should even be able to create text, and we have also more stuff. Let me show you something from Leonardo AI directly. Leonardo has put together a nice image guidance. So here you have a nice overview of this image guidance. You can see normally we are able to take this control net to the next level with depth, even sketch, post normal pattern QR code, inert, head and text image input. Not everything is included all the time. If you scroll down a bit, you see how every single thing works. Depth to image, you've already seen this. It works like this. Etch to image, here is a nice example from Leonardo. Linart, this is not included right now, but should coming soon. Then we have the sketch to image. Sketch to image, we will take a look at the AI Canvas because in the normal interface, this doesn't work at least right now. Post to image, you've already seen this normal map, This is right now not included. I think this will get included soon. Same thing with pattern to image, this was included. Got removed, and I assume this will get back in no time. QR code to image. This was also included for a long time. Right now, it's excluded, but you can make your own QR code if you simply use image to image. So yes, this works right now. Just use image to image for these or depth to image. You need to play a little bit with these. But exactly this tool right now is not the viable anymore. And the text to image right now seems to be not optimal. You see, yes, they bring a lot of new stuff all the time. If you scroll down, you can see what thing you can do in what model. You can do nearly every single thing in stable difusion 1.5 or the stable difusion version two. And of course, in other version, you can do just some of these stuff. So make sure to use the right thing. And that's basically it. This is everything you need to know about the normal picture generation inside of stable division and Leonardo AI. You know all the basics, and of course, in the next section, we will make a deep dive because we have the AI Canvas, three D text generation training our own auras and much, much more. So stay tuned for the next section. But before you go into the next section, I have just a tiny bit of homework for you. Come on. Just go in Leonard I. Search the right model. Maybe you can even just remix something from the gallery, and then play a little bit with prompts, negative prompts and include some of the stuff like alchemy, if you have access to that or use the things like image to image or even the control net. Like I said, if you have access, just try this out. I think this is a cool homework. So don't let me down on these. 93. What this section is all about: In this section, we will make a deep dive into Leonardo AI, not just like the picture generation but also editing of the pictures and much, much more. You will see the real time Canvas, the real time generation, the Canvas editor or the painting and out painting, if you want to call it this way. The texture generation, how to make videos with Leonardo motion, and of course, also how you can train your own AI model inside of Leonardo AI. That's the easiest way how you can train your own stable fusion, Laura. Stay tuned for the next lectures because all of this is awesome. 94. realtime Gen inside Leonardo AI with SDXL Turbo: In this video, I will show you the real time generations. These are awesome. These are enormous fast, and also this video will be fast. So if you go on the left side right here, we have the real time Jen. Yes, we have seen image generation and the real time Canvas, I will show you in the next video because the easiest thing to start are these real time gens. And if you simply press on these, we can create our pictures in real time. That's no joke. We can include elements, so these are the auras, and we simply leave them all at the default settings, you can play with these if you want. We use a random seed. You already know what you can do with the seed. Then we can also use different stuff like here. You can make ems cinematic and much, much more. I just use dynamic because I want to show you what we can do right here. If we start typing, we will get immediately results. This is our A and we get this. Then I'd like to say, for example, A, T, If I type in T, we get different stuff. If I delete it, and we have just this first letter, we have this. Yes, this is really so fast. If I include the next letter, we get always other outputs. On the street, mountain, whatever you like, if you change it, also the picture will change. Mountain right here, you see this works really that fast. And if you like to do this, you can also do something like this. Woman 20 years old. Then you can simply see how she gets older. 30, 40, 50, 60. Of course, if you want to do this with a fixed sat, you can also do it. So we use not a random seed, but I want to use the seed of one. Of course, you can also start them young. So two years old, ten years old, 20, 30, 40, I hope you get it. You simply type in what you want to see and you get in real time the things that you can create. And then you can also do this with like Laura's and much, much more. Simply increase the Laura, decrease the aura, and you see immediately what you get. Right now, aaura of zero with this anime style, 0.2, anime style, 0.5, and now a full blown me character. And now we go back like 0.1. Kids illustration, full blown kids illustration. You see you get messed up results if you use too much auras, and if you use the style too strong. So just simply play with all of this a little bit around that you will get what you can create. Coloring Book include also this, and now we get a little bit of a mess. I want to exclude this and the output gets better. In this video, you saw the real time generations. These are fast. Try them out. 95. Realtime Canvas inside Leonardo AI with SDXL Turbo: Besides the real time generation, we have also the real time Canvas, and these are even cooler. Come with me and take a look. We simply press on real time Canvas right here, and then we are in this interface. So right now this is empty. I will exclude myself so that you can see everything better. In the left corner, you can exit all of these. Here, you can simply add an image. Select stuff, delete stuff, use a brush. Use the brush size, undo, redo and delete everything. You can draw or in paint. And then the next step you need to do is to type in stuff right here. For example, as soon as I start typing, for example, the letter A, we always get outputs. The output is in the right corner right now. Now let's just say a dog. You see, we get immediately stuff here in this pictures. And of course, I can delete it and make new stuff and we get new stuff every single time. Now, the coolest thing is, we can also paint on the left side. Let's just use this brush. I want to make a little bit of a smaller brush, and let's just make something like this. Maybe we get an eye of the cat. And bam, there you have it. So you can simply draw stuff right here and you will get stuff on the right side. And this is really, really fast. Let's just make a second die right here, and maybe a mouth that is open. Okay, this is not perfect. What if we use some teeth right here. Okay, this is maybe a bit stupid, so we can also undo stuff if it doesn't work. So instead of a cat, let's just say this time a house. And now you can simply make your house how you want. So I decrease all of this, and then I can make my house something like this, for example. You see, we get the stuff here. And then let's just say I want to have another house right here. So we have two houses right now. So you see how fast all of this works. And then I said house and maybe we include some sun, and then I simply want to draw the sun. We can also do that. So you can see how fast and easy all of this works. Yeah, now this sun is maybe not optimal. So now we get better stuff. So you can see you can create every single thing really, really fast, and you can simply play with all of these. So this works really fine right now. So I think we get good output right here. And of course, you can do this really however you want. You can also do something like this output to input. And then you change this and now you can also make your new stuff right here. And of course, you can create whatever you like. Now you can also make something like let's just see if we can include stuff right here. You can see we can make this a lot more shiny if we simply do it this way. All of this works relatively cool, and this is not all because you can also do in painting for the in painting, we simply press on these. And here you can see how this works. So we press confirm. So we can simply use this in painting, and now we can add stuff in our pictures. Let's just try if we can add something. I want to make, for example, something like blue colors here. We can use brush. This works for me. Now we need an painting prompt. So what we want to paint, let's just say sky, and then we want to have the sky a little bit more blue. And for that reason, I simply also make this a bit more blue. Okay. And you see as soon as I start to make some stuff right here, we can get a lot of blue sky. So I think this is cool. I think you should try this out. You can also do stuff like bringing your own pictures. If you press on that image, you can either upload the or select an image. I just use my image right here, and I press confirm. And now I can do in painting and drawing and all of these. So right now, I just want to do the in painting, for example. I use this brush and I can do stuff like I can add something right here, for example. And I need to type in in the prompt, what I want to add. Let's just try neck tie. And it worked. So you see how easy this gets. Yeah, this is really, really phenomenal. Of course, you can always undo this. You can delete the stuff. You can also do stuff like let's just try this Superman. And now we make a bigger brush, and I want to make me as Superman, and you are able to see this in completely real time. So you see, I get transformed into Superman brush stroke by brush stroke. This is enormous. Just look at the muscles I think. I should train again a little bit more because I would love to look that way. So all of this looks really, really cool. And of course, if you want to have a prettier face, we can also do this. So we can make my face a little bit prettier. And every time we make a little brush stroke, we get prettier and prettier until we are completely superman. So this is really, really awesome. And of course, if you press back, you can always delete this stuff. Let's just add my picture one more time again because we can also do this eraser. So let's just assume that you want to have a completely black thing right here. I think this is reasonable. So you can totally do this. And then if you don't like it, press back. Of course, we can also press drawing right here, and we press confirm. Now, if you use drawing, none of this will work. So you need to do the in painting if you want to edit your pictures any further because It simply not works that good. So if I use man right here, we have a little bit of reference picture here from me, but none of this is really, really that good. And now you can simply delete or include stuff that you like. Let's just delete something right here, and you see we get blue stuff in the background. So as soon as you use draw, your input image will not have that much of an impact on this thing right here. And then you can always press output to input. And here you can't draw any further if you want, and you can also press the in painting and add here different stuff. Like, for example, blonde hair. So we simply do here the painting. I brush this here, and then I do blond hair. And now we have blond taar. So you see how easy all of this works. And I think I need to make myself also blonde one last time because this works so fast and so cool, and I think I want to be a bit cautious, the brush not too big so that we get nice little blond taar for me. Yeah, maybe a little bit more blonde. So you see how nice all of this works. Maybe also here some blond tears. We can also do it really, really enormous and make this blond tear bigger and bigger until I get the complete lion. Maybe I can even include some hair right here to see how I would look with more and more blond hair. So you see more and more of the picture gets recreated and my head looks a bit funny now, and that's it for this video. This was the real time Canvas. You can play with this a lot. And to be honest, I see really a lot of architects play with this stuff right now. I think this is somehow the future. If you are really, really good with this Canvas, you can make a career out of this. Maybe not right now, but this will get better and better. I saw a lot of stuff going completely viral on Instagram and YouTube. There are people out there that make really, really awesome stuff with this in painting and out painting. I myself, I am not an architect, so I think something like this was everything that I can do, so don't worry. I will not take your job if you are an architect. But if you are an architect, you can do amazing stuff with this tool. And thanks to this course, you know how to do it. Just play a little bit around and mess with this stuff. Stuff is really the future. Just imagine what you can create in really no time. If you need to make creative stuff, this is the thing that you need to try. 96. Leonardo AI Canvas for Inpainting and Outpainting with Stable Diffusion: We have the possibility to do in painting and out painting in Leonardo AI with stable division, and we can even use different models. This works exactly the same way as in the Dali platform and in adobe firefight, but it's in my mind, at least the best option of all of these. Everything you have to do is to press on Canvas editor right here in this interface. And here, you see, we have a lot of different options. In the left corner, we have right here the exit button at the top. Then, of course, we have the pan tool. We can simply pan everything around if we want, and we can of course, zoom in and zoom out via our mouse. Then we have this leg tool. So same stuff as in Dali as always. We can draw a mask, a race. We can even add catch text. We can upload image. We can delete everything, and we can doload everything. Of course, we can do stuff. Then if we go here on this side, this will always change depending on what you use. Then you can, of course, always zoom in and zoom out on this button. And this right here are the keyboard shortcuts. So if you ever want to use shortcuts, just look at these. And here you get also some infox. If you press on these, Leonardo AI will give you a quick guide what you can do with this stuff and how every single thing works. But don't worry because I will show you everything. On this side, you have, of course, your credits, and here you can say what model you want to use. You can use a lot of different models. If we work with stfsion 1.5, we have our normal model. Then we have the Canvas mode. If you press on the canvas mode, you can simply see for yourself what you should do. Do you want to make normal images? So text to image? Do you want to do in painting or out painting? Do you want to do image to image? Because also this is possible. So transform the underlying content subdly or completely. Equally versatile for touch ups or total re imagines, and then you have sketch to image if you want to include some sketches. Let's just press text to image just to start so that we have something to include right here. And of course, you can use the number of images that you will create. You can include photo realismus, alchemy, prompt magic, and so on, but just if you use text to image. Now, if you use the in painting, for example, You don't have all these models. But right now we want to make our image, so we use text to image. And for the text to image, we have, of course, also all of this stuff. I use the photorealismus, for example. Also alchemy is activated. We can use different image dimensions. And here we can do every single thing just like in the normal interface as soon as we create pictures. I start with a really, really simple prompt right here. Man on the beach, and I press generate and then we will create a man on the beach in this interface right here. And here we have our pictures, and always remember you can use different stuff right here. And let's just see what I like the most. I think I like also this first picture. So also this looks relatively nice to me, and then I simply accept this picture, and then we can and will start to edit this picture. But the editing is for the next video. And of course, if you don't want to edit this picture, you can simply press on upload images. So you can, of course, edit every single picture that you like. If you press on these, You can upload a picture from your computer from previous generations, or, of course, also from the community showcase. So you can upload everything that you want and you can also merge pictures together if you like. So you can make, for example, two, three or four images and then merge everything together, and you will have fun. In this video, you have seen that we can use the AI Canvas. In the AI Canvas, we can do a lot of stuff. In this video, you have shown the interface and also that we can create pictures inside of this AI Canvas. The next step is, of course, to edit these pictures, and I will show you the editing in the next video. 97. Inpainting in Stable Diffusion with Leonardo: In this video, we talk about the painting and out painting with stable difusion. And of course, we use the AI canvas of Leonardo AI because this is at least in my mind, the best interface to painting and out painting for stable difusion. You already know how all of this works. So basically, we created a picture. This picture is Man on the beach we did this in the last video. And now I want to show you how we can edit this picture. It's really easy. Of course, you can always zoom in and zoom out so that you can see this better and use the pan tool to make all of this seamless. Then we go onto the Canvas mode, and here we need to change. And I want to exclude my face so that you can see all of this better. We don't use text to image because right now we need in painting and out painting. So I press on these right here. Now we have set in painting and out painting. The out painting is also enabled. So if we don't enable out painting, we have just the possibility to edit this picture, and if we include this, we can also make stuff outside of our picture. So we can, for example, make stuff also right here. If we include this out painting, and if I don't use it, we can't do this. Then we have the paint strength, and of course, always the same stuff. If you don't know what something does, you can go with the mouse on this question mark. So the painting algo looks at what already exists in your image to better edit or place new content. The higher the strength, the more it will diverge from the original. Then we can use the numbers of images. So how many things we create. I just use most of the time four, so this is the default settings. The image dimension, you see, we can use different stuff, different respect ratios and different resolution. And if we go down here, you can, of course, also do this manually, so you can increase or decrease specific parts here. And you have also the render density and the mouse says everything, decrease the size and increase the pixel density of the generation context. Of course, if you increase this, your generation frame will get a lot smaller. And if you decrease it, your generation frame can get a lot bigger. The guidance scale just leave it at the default settings because if not, you get this right here. So seven or eight works most of the time perfect. The recommend that size, I would simply let this include, and we don't use a fixed seed because it's better to let stable difusion guess the seed that we used. You don't use oiler. So I think Leonardo is the right move to use right here, but if you really want, you can play with these. Like I said, Leonardo is my favorite. After Leonardo, I use oiler sometimes, and the other things I don't really use them. So we use Leonardo. If you ever mess something up right here, you can press set to the forts. And then we are ready to go. So right now, I simply want to do in painting in this picture. Let's just assume I want to add something into this image. So we can add, for example, Let's just say a dog or a cat, or whatever. We need to make sure that we have the in painting strength relatively high, and we can also increase include this. And then we have two different options here. Either draw mask or the eraser. And of course, it's different what you want to do. And Leonardo tells us what's the difference between masking and erasing. Masking an era retains some of these objects and styles underneath, making it ideal for minor touch ups and raising an era completely removes unwanted parts or prepares for replacements. So if we want to make a dog inside this picture, we need, of course, not mask but erase our stuff. So we erase something like here and we try to make a dog here. I just erase all of this, and of course, we do this really, really simple. I just type in dog on the beach. So this is nearly the same prompt, of course, and I hope that we get something cool. And of course, we have here the models. You can always use the model that you think is optimal. The standard model is always stable diffusion 1.5, but you can also try absolute reality for this specific example. But like I said, right now, I just want to go with the standard model, and you can probably should play a little bit with different models because different models will generate different outputs. And this right here is somehow okay. The other pictures are a little bit of a mess, I think, at least, especially this picture is a complete mess. If you don't like your pictures, you can also cancel them. You can tell, for example, back of a dog on the beach. Let's just see what we get here. I want to don't see the face of the dog. I think we will get better outputs if we exclude the face. And I played a little bit with the prompt because most of the outputs are not optimal, but I think this is somehow okay, so I need to accept this. Of course, if you want to of course, if you want to add completely new stuff, this is not so easy. So you need to run this a few times, and also this is absolutely not perfect. But this is just to show you how all of this works. Now I want to show you what the mask can do. So we use a mask right now and we can simply mask certain stuff out here. And of course, you can always increase or decrease this brush size. I think I want to mask out this things right here, and I want to have ocean and not the stones right here. So I simply delete all of this, and of course, I can also leave this empty. For that, I simply press space one time, and then I press generate, and this will get masked out. And I think we will get super good creations at the first try because normally this works really good. By the way, right now, I use the absolute reality model, because this worked better than the stable diffusion 1.5 model. You need to simply play a little bit with this, and there we have it, so it is masked out, and it is at least in my mind, completely seamless. I think I like this the most, so I accept this. Of course. You can mask and redraw and do whatever you like. You can also add, for example, a few birds right here or mask the hair and make the hair blunt and much, much more. That's the basic concept of this in painting. So in this video, you have learned how the in painting works with stable difusion in Leonardo AI. You need to use the right model. You can simply mask or erase certain stuff. If you mask you can simply mask the stuff. With the eraser, you can delete and make new stuff, and that's basically really, really cool. And in the next video, I want to show you the out painting because the out painting is really, really similar. So we merge two pictures together in the next video with the out painting. 98. Outpainting and marching Pictures with stable Diffusion: You saw the inpainting with the eraser and the mask. In this video, I want to show you the out painting and how you can merge two pictures together. I think you already guess how we can do this. Step one, we do the out painting. Of course, we need to use our right model. We need to include out painting right here, and then we can simply move our generation frame in the direction where we want to out paint. Let's just say we want to out paint in this direction. And right now we can and probably should also use the eraser or the mask. I can simply mask a few parts right here so that stable difusion can do better overlaps right here. I simply mask a little bit. This is really, really better because if we don't do this, stable division needs to create the new pixels at a hard edge, and this doesn't work that great. Right now, you can include prompts, but you don't have to include prompts. We can even run this without any prompt and then stable division will simply guess what in the picture needs to be. Maybe we get just a little bit of ocean and a bit of sand. But it's also possible that we include humans or something like that. Okay, this works perfect. We have just a little bit of more ocean and also a bit more sand. And you can simply see what we have on these pictures. The first picture is not bad, bad. The second is nearly optimal. The third is okay, and the fourth is awful, I think. So we accept the second one. And now I want to merge this picture together with another picture. We can, of course, create a new picture or we can search for a picture that we already have created. Or we can do other stuff. We can also upload two other pictures. So we don't have to merge this picture together with another picture. I think we do something completely new because you already saw this picture. So I search for new pictures, and we do this, of course, here with upload image. I like to search stuff from the community. I think I want this bird and that bird, and these two birds need to be friends. So first, we upload this bird. Then of course I do the same thing again. So upload image from community, and now I use the second bird. And this looks really, really cool. I need to zoom out a bit. I think I can and probably should also delete the other pictures that I have made. So we simply delete these pictures right here completely. And now we can use our other pictures, and I think we need to merge these pictures together. I can make them a little bit bigger or smaller or however we like. I think something like that would work fine. They should be friends, so they should ideally look in in this direction. We can simply play a little bit with all of these. So I hope you get the concept. Now, I think this looks perfect right now. Of course, you can always use different tools. You can zoom in and zoom out. Maybe it's a little bit too small, so we can make it just a tiny bit bigger. Right now, I think it's perfect. And to merge these things together, we can use the out painting. So we simply need to use in paint and out paint. We go with the generation frame in the right spot. Of course, we should make this a little bit bigger just to make sure I will use it right. Here, for example, now I use draw mask or the eraser. It doesn't really matter that much. I make this a little bit bigger, and then I simply go over these edges to make your seamless thing right here. Because in one picture, we have a lot more diamonds and stuff, and for that reason, I simply want to delete a little bit of these. And then we let stable difusion completely guess. So of course, not a house on the beach. We simply press generate, and then we see if stable difusion merges the two things seamlessly together. I think this looks cool and what the other generations, so maybe Maybe the first one was already the best one. I think I want to use this and I simply press accept. And of course, you can and probably should use other models. So if you press back, if you don't like it, we can simply use other models, so not the stable sion 1.5, but something like the absolute reality. I think this makes better generations with this specific example. Now, this looks good for me, and now we simply draw a mask one more time. We simply mask out a little thing right here and also there, and we create our new stuff also here. And of course, if you don't like something, you can simply delete it with the eraser or with the mask after your generations. Let's just assume you don't like this messed up thing right here. We can simply go with our generation frame over these and we delete it completely. I think the first one is really good. Also, the other ones seem relatively good to me. I love the second one the most. And of course, this is emerging together, and you can also out paint every single thing. We can simply go right here. We can work with prompts or without prompts or whatever you like. We simply press generate, and then we will create new stuff right here. It's always important to let at least 30% of the generation frame inside the picture. Stable division absolutely needs this to see what's going on in the picture. Stable difusion needs to read the pixels so that stable division can generate similar stuff. I think the first picture is awesome. The second one is a mess. Third one is also awesome. The last one is not that good. So I think the first one is the best. So I accept the first one, I do one more little tiny thing right here. And remember, we don't even use a mask right this time. Right now, I will use a mask because here we have two edges, and I think two edges is maybe a little bit too much. So we do this right here. Maybe I mask out also this. We press generate one last time, and then we have also out painted this picture in perfection. Then I will delete this last little thing right here, and then we have our complete picture. So this looks perfect right now I draw a mask one last time because I want to delete this stuff, this looks really a little bit messed up. Also, this is a little bit messed up. And maybe also this looks just a tiny bit like a finger, so I want to exclude also this right here. So we mask this thing out, and then we have our stuff. And if you have ever something like this right here, we can always use this select tool. So here are old generations, and we can simply move them to the side. So you can see here is a little bit of old out painting stuff. We can simply move this to the side and you can also delete them if you want. If you press select, once again, you can delete this little stuff right here if you don't like these. I have moved also this frame to get to the side, and now I am relatively sure that we get good outputs. So yes, this is way better. The four is perfect, so I accept these, and I think we have right now a really, really beautiful picture. What do you think? In this video, we saw that the out painting and the merging together of pictures works really, really perfect and seamless. Stable diffusion is the best tool for out painting and in painting, and especially in the interface of Leonardo AI. And this tool has more features. One tool is, for example, the sketch to image, and about the sketch to image is the next video, so stay tuned. 99. Image to Image in the AI Canvas of Leonardo: Before we go to the sketch to image, I want to show you the image to image. The image to image is really, really cool. So of course, you can use the image to image if you go into this mode and simply press on image to image. Right now, we already have a complete image right here, and this is completely the same concept as image to image in the standard picture generation. Now, we can use always as select tool, so we need to select our generation frame. And if we go with our generation frame four example over this bird, we can create other images based on this image. So we can simply, for example, type in bird right here, and we will create an image that looks somehow similar to this thing right here. So it's simply image to image but in this cool interface. If I press generate right now, we will generate a bird, The bird will look relatively similar but not the style. Just the general feeling of the picture will be similar. Now you see how the bird gets transformed, so I hope you see what I mean. This is really, really cool. And of course, we have four different outputs right here. So we will simply create something that is somehow similar. If I accept these. Now, we have this picture right here. You know that this picture is not perfectly suited for this place. But of course, you don't have to use it at this place. We can also use as a elected tool and take our picture, for example, in another place right here. And I think this is really, really awesome. So we can use this image to image. Go in different places and create new pictures. Of course, I can use the generation frame right now. Maybe I can go over this bird and make a similar thing. I simply press create one more time again to create a bird that is somehow similar to this bird. Of course, I should probably work with better prompts. So we can say explicitly how this bird should look and much, much more. And of course, we can also increase or decrease the input strength, just like you know it. It's exactly the same thing as the image to image in the standard generation of Leonardo AI. So if we increase our input image, for example, at 90%, we get nearly exactly the same bird as we have as our input image. And if we have it only at 0.3, just like we have it right now, we will get slightly different birds, of course. And then we can simply see what bird is cool for us. I think I like every single one of these. We can accept one of these. And of course, you can also take the bird into other places. So for example, you can take the bird right here, and then you can merge these two pictures together if you like. And like I said, you can also make this bigger and smaller so the generation frame. Let's just increase this a little bit. You can also go for example right here in this corner, and then we can also increase this stuff, so increase the input strength at 0.7. We use this bird, we press generate, and we will recreate something that is really, really similar to this picture. It will be 70% similar. So similar style and every single thing will be much, much more similar than in the previous generations. And there we have it, so we can simply accept and you see this is really nearly the same picture. I can use this picture and you will see. It's just basically the same stuff because, of course, we had a high weight on the input, and of course we recreated nearly the same picture right here. And now, if you don't like something you can, for example, delete the stuff that you don't like, you can delete every single thing that you don't like from the pictures, and you can simply download the stuff that you like that you love or whatever. And if you want to download, for example, just this tiny picture picture right here, so we have this picture right now, we can simply press download artwork and then it will get saved on our computer. In this video, you have learned that image to image works also in the AI Canvas. You can play with the strength of the input image, and the input image is always the generation frame that you point over your picture. You can play with strength with prompts and much, much more, and you can do whatever you like after your generations with your pictures. And in the next video, I will show you really the sketch to image because right now we have the other stuff covered. 100. 3D texture Generation with stable diffusion, make Textures for your 3D Printer: In this video, I want to talk about the three D generations inside of Leonardo AI. This is one of the coolest features. And if you have a three D printer, you can take all of these, and then you can print your own objects. Yes, you can make physical objects out of a ODT file, and then you can print them. Everything we need to do is, of course, to upload the ODT file into Leonardo AI and then we type in some prompts and we will get cool results. So step one, you need the ODD file. You can make them yourself. If you are a pro. If you're not a pro just like me, we need to search them online, but we can get them for free. The website is called 33d.com, what a name. And here we can search for OBGFiles. If we simply press on these, we can search every single thing that we like. So let's just assume that you like Bugati or a male base or Iron Man. You can simply search for the stuff that is free, and you can download it. And of course, if you need something, especially, you need to pay for these. They have really a lot of OBG vials and you can simply download all the stuff that you like. As soon as you find something, you need to simply press on these and then download it. So this is free and this has a lot of downloads, so you can simply press on these, and then you can download it. Now, after you have downloaded your small OBG Val, you need to press on Dexter generation inside of Leonardo AI. I have already uploaded a few models right here. And if you don't have anything right now, you simply press on upload new objects, you give this a name, and then you add your file right here per drag and drop. But you see right now, I already did this, so I have a cull and I have a man, and then you can simply press on this and we can generate really cool output in all of these. You will be in this interface, and normally this stuff is completely white, but I already did this with this nice little ODT file. I simply laid one prompt over this. Now I will show you how all of this works. Like I said, normally, this is completely white. You can simply go over prompts and see what this stuff does. Your prompts will be used to generate the texture sets. So you can simply include right do you want to see, so you can use a single word or a full sentences. Prompt engineering here is not that important. I think I worked with the prompt diamond. Then we can simply create a preview. If you press on this, we pay only five tokens to create our preview. And like I said, I already did this. So this is the preview of the diamond skull right here. And if you don't make any preview, you just have your white skull here. You can also include negative prompting if you like. And if you think this is perfect for you, you simply press and create full texture, and then you get your full output. Before you do that, I would really recommend you to just make this small little preview so that you can see what you can create. So here, for example, you can see some of the diamonds. And before we really create everything, just look at what you get, because I think this is important. You can also go on settings. Because on the settings, you can adjust every single thing. So here on the adjustments, you can increase the mesh adjustment and the Zoom. So you can simply increase or decrease this stuff. And then you can also press on assets. Because you can use a different sat. So if you like, for example, a really cool sat, then you can see the prompt. I used creative bin crystalkull and not just crystal alone. So excuse me. So this was my original prompt. I don't use any negative prompt engineering and your generation assets here, you can see what we got, and we can, of course, also download all of these. So I assets, you can fine tune every single thing. Then you can go back on generations and create your full texture if you think this is nice, and I will do this right now, and I will see you in a few minutes because this takes most of the time like half of an hour. And here we are again. So this is my full skull. So this is my creative crystalline skull, and you can see we have our whole three D object. You can see we have really included the prompts in every single detail right here, and we have a big creative skull with a lot of diamonds, and I think all of this looks really cool. And you could really print this with a three D. Printer. So I think this is completely awesome. So you can also zoom in and zoom out really, really nice. All of this worked enormous good. So I think this is a really cool tool, and you should try this out. And after your generations, of course, you can simply press on assets and see everything for yourself. So you can download all your assets or you can also download just the OBGle. So in this video, you have learned that we can even make three D objects inside of Leonardo AI. Everything we have to do is to download the OBG file. We upload it in Leonardo and we type in our prompt. This takes about 20 to 30 minutes. Sometimes it also takes just 3 minutes. You will pay a lot of tokens for these generations, but after this, you can also use this for three D printer. So this is completely awesome. Just search for a OBG file, for example, on 33d.com, and you can upload it and you can play with this a little bit. Of course, if you can create your own OBG viles, just feel ready to go. So have fun with your Freed printer. 101. Sketch to Image and adding text in Leonardo Canvas: In this video, I want to talk about sketch to image and also about including text in your artwork. All of this works really easy. All of this is really cool. I already uploaded a picture from the community showcase. So I have simply stolen these. Then we press, of course, is always on the Canvas mode. Now, we use the last thing, sketch to image, transform your sketches into a generated images, either freestanding or incorporate with other content. And you see right here, we have Monalisa with a little bit of beard. We use this and now we can do our sketches. And of course, to include our sketches, we don't use the draw mask. We don't use the eraser, but we use sketch. So we press on this sketch, then we can go with the mouse over our stuff, and we can make this bigger and smaller. If we can make really, really small stuff, for example, this thing right here, and we can draw things. I think this really, really awful drawing, and let's just see if we can make this really, really awful B. This is not likely how you do it. You can do this a little bit more professional if you like. But right now, I simply just want to type in here, for example, B, and then we see what we got if we get something. And of course, you can also go over this right here, and you can simply use different colors. So if you go for example on this orange thing right here, you see we have our orange colors or we have our cream colors, and so on. And now we need to use our generation frame. So I simply select the generation frame. I make the generation frame a lot smaller because this is way too big for our small thing right here. We go over this stuff right here. I use the prompt B. We have made a really cool painting right here, and I simply press generate, and then we see if we get something out of these. And you see right now, this is completely vanished. I wanted to completely show you this because a lot of people make this error. Why this doesn't work. Right now we cancel this. We go back because this was completely not good. We need to go back, so we simply need to delete all of these. Now, if you use sketch to image right here, you press, of course, on sketch to image but. You need to increase the strength of the sketch so that you can draw really, really good. You need to completely exclude every single thing that you want to exclude, and you need to take a color for these. So if you want to make a B, take, for example, the yellow color, and now you can make your B. And of course, you need to make your B complete. So you need to do something like this. We need to make our whole B so that the model can make something out of these. So maybe a B would look something like this in a sketch. Of course, this is awful, but the generation will fix this a little bit. And now we can come with our generation frame and make something out of this. This is the key right here. Of course, you need to make this a little bit bigger, and we go with the generation frame in this angle and we press generate and then we should get a nice little sketch out of this. And there we have our nice little B in the color that we have included. So we press, for example, except. We have really a funny B right here in this picture, and of course, you can add other sketches. Let's just make a little beard to this girl. We take, of course, our sketch once again. Now we take your color. I think a brown color should work somehow with this. Let's just try if we can find a good color for the Let's just make a red or brown color for the beard of this girl because I think that should work a little bit better. So maybe this is a good color. And remember, if you want to have the exact color, you simply press on these and the color will get included automatically. Now we press on this sketch once again, and now we can make our sketch. So I think she should have some beard right here. Maybe down until here. So we want to make a nice little beer tier. We also include the tier just a little bit. So she should get a little bit of an orange beard, and I think she would look awesome with this beard. This is how you make sketches out of these. Maybe even a bit higher, so she should really have a good amount of beard unlike me. And we should also use the stuff that we want to make. So for example, beard. Now we can press generate and now we should be able to get a really good sketch out of this. And there we have it and I changed the prompt a tiny little bit. So I changed it to orange beard and harry, and we accept these because I think this is a nice little sketch. And of course, you can create these sketches not only in pictures, but you can also make them completely separately. Let's just increase here this output just by a tiny bit. We take our generation frame right here. We work, for example, with the prompt house. We take our sketch right now, and now we can make a house, for example. So we can make something like this. You will see we can create a nice little house with these nice little sketches, and of course, the house will be awful because I am a awful painter. This right here is my nice little house, and I suppress generate, so we leave the color and every single thing just at the default settings, and we see if we can make a house out of these or not. And there we have it. So this is basically our house. If I accept this, we can download it and we have a really, really cool picture or what do you think? This is how we can make nice little sketches in your images. And if you go with your generations outside of this stuff, the sketches will always be right here in this little corner. So we can always delete the stuff that we don't like. And of course, if you like the stuff, you need to leave it. But I don't think I did a great job with the sketches. And now I want to also show you that we can include text with one simple click. If we press on these right here, we can include text. You can also see right here that we can use different colors. So for example, let's just take a little bit of a black color. And of course, we can also use different text. For example, this impact text. Now we simply press on the stuff that we want to create. And now I simply write in my nice little text. So for example, no catches. Of course, you can also change how big the text is and much, much more, you can simply increase or decrease here, the strength of the text. So how big the text is if you want to make this a little bit. Faster, you can do it right here. And then you have your nice little pictures. The sketches got deleted. But instead of the sketches, we have our text now included, and you can move your text around. So editing your pictures was never so easy then in Leonardo AI. I think this is like a really serious concurrent to photoshop. In this video, you have seen that we can include sketches to our images. I did an awful job, and because of this, I deleted every single sketch out of my picture, and I just included some text that says no sketches. But I am 100% sure. You will make better sketches. And that's basically also what I would recommend you to do. Just include some sketches in your pictures if you think they are funny. And then you can also include some text. It was never so easy to edit your pictures than it is right now with this division models. Especially Leonardo AI with stable division in this interface is enormous, easy to use. We can make completely different pictures in no time whatsoever. I really think I honestly think this is a good concurrent D Adobe Photoshop because we can do a lot of stuff in Leonardo AI. So try all of this out. All of this is awesome. What do you think? 102. Train your own Stable Diffusion 1: In this video, we will train our own small stable diffusion model. Training our own models is really, really cool. We can train models on our own face, but we can also train models on other stuff. Let me just show you one model that I have trained. This, for example, is a model that makes close ups of insects. So you can see I have created a closeup of a B and a close up of a spider. Now you may be think why this is special. Just look at the prompt. B Zc and Spider zt. This is the prompt and I get really awesome results. And the coolest part of all of this is it just causes me one single token to generate such a picture. So if we do, for example, one more time, B insect, we will create a really, really good and standing picture because I have trained a model on a lot of close ups of insects. This works really that good, and we don't need any prompt engineering whatsoever, because the model is simply trained on a lot of outputs that look similar to the stuff that we want to have. And because of this, also the generations are so cheap. We don't need any extra fancy stuff. We don't need nothing. Now I want to show you how we can train a stable vision model on our own face because in Leonardo, this is really, really easy. Step one, we press on training and datasets. Here you have your job status, so if you trained a few models, they will be saved right here. You can see right now I have the five models. And then you need to press on your datasets. You can see right now I have seven different datasets, and I just messed a little bit around with different inputs and much, much more. Now, I will show you how this works because all of this has a few tricks. First, we press new dataset. Then we need to give this dataset a name, for example, Arnie in the AI. Because I want to make a model of me, then the dataset description, me in the AI, for example, then we press create the dataset. Now, here we need to upload our dataset images, but here it's important that we use the right dataset. If you scroll down a little bit, you can also search on the personal feet or on the community feet. So you can even train a model on pictures of other people. But the most important stuff is that you use the right resolution. Now, you will get the best outputs if all your pictures are in the same solution. And the best thing is an output by 512 512 pixels. Yes, you can also train models that make other outputs. But for right now, it works best if you use pictures with 512 512. Now, if you don't have pictures of you in 512 by 512, make some pictures, and then you need to crop them. And right now, I will show you where you can crop your images. Berm.net. This is the website that you should use. Here you can simply upload pictures. You can upload up until 30 or 40 different pictures, and then you can simply include the width and type that you like. So on aspect ratio of one by one is perfect, and then you can simply crop all of them and download it. It will look something like this. Now, here, I just uploaded a few examples, so you can adjust everything that you want. So we can simply turn this around and you need to make your pictures square. So 11512512 pixels is perfect, and then you save this as a zip file and you are ready to rock. I already did this with the dataset that we need, and now I will show you how this looks. This right here is the training data set for our model. And you can see, I have simply every single picture in one by one. This is important. So every single picture is one by one, 512 512 pixels. I have a lot of pictures of my face. I have some pictures of my body, so we have three pictures of my body. And it's important that we use different background. So you see black background, white background, blue background, red background, and so on, and I am not completely similar in every single picture. The more angles you can include, the better the output we'll get. So it's really makes a few pictures with different background, and you crop them in 512 by 512 pixels. Now, these pictures, we will upload them. And you see it right now, I have 25 images. You get the best outputs if you use 10-30 different pictures. So now we go back on Leonardo AI. And here, you can basically it, drag and drop files to upload or click here, and I simply want to drag and drop my 25 images in here, or it's maybe just 24. I don't know it exactly. So we have 24 or 40 images. And Leonardo tells us in another article that something 10-30 is perfect. Just start with 15 pictures if 20 or 40 is too much for you. So you need just a view pictures. All of this is important and it should be. 512 512 pixels? Because if you don't do this, the output will not be that good. And make sure that you don't look completely similar in every single picture. So you see I have some ugly pictures, and maybe every single picture is ugly, but I think this is completely okay. Now we press train model. And here we need to use our stuff. So what's the model name? A rne in the AI. I think this works. So Arnie in the AI. Then the training resolution is, of course, 512 by 512. Now you can also use this output, but you have cropped your pictures in five 12 by five 12, and to be honest, this works better right now. Then we need to use our category. So what should neural building? Character environments, fashion illustration, game item, graphic elements, photography pixel art, and so on. But I think we should use characters because we make like characters. Maybe if you like, you can also use illustration if you just want to make illustration. I use characters right now. The training settings will be adjusted to suit. Then we need to make a small model description. This is also important. So what we make? Let's just say me in the AI. Then the base model stable difusion 1.5, and you see it also right here. Stable difusion 1.5 is recommended. We can also use stable difusion 2.0. 2.1, but this doesn't work that great, and it's may be better if you want to use higher resolution. So we use stable division 1.5. And now we need to start with our instance prompt because we always need to include this little word as soon as we make our prompt. I just want to make like caracter. And we can also be more specific a character of Arnie. So this is our instance prompt and this we should include in every single prompt. This model is not safe for work. I think this model is safe for work. And then we start training. We have ten remaining models. So I press start training right now. Training in process, we will e mail you when it's complete. Depending on the size of your training chop and server load, it may take anywhere from 30 minutes to a few hours. And you simply press view chop status if you want to see this for yourself. And you see right here we are in the training right now. The status is processing, but most of the time this is really a lot faster than Leonardo tells you. Most of the time, this is done in like maybe five to 15 minutes, and if not, Leonardo will simply send you a mail. So I see you as this is trained because then we simply test our model. And here we are again, this took about 10 minutes, and now we have our model right here. So you see it's done. I can simply press on this model, and here we can see all the details. So I close this right here. And if I go on fine tuned models, I will see this model right here. So it's the first model, RN in D. I can simply press view on this model, and then of course, I can also create with this model. And here, you see, we have this model, and we have always the instance prompt. So a character of Rn. We should always make a prompt where we include this right here, and you can also see it if you go over this question mark. We highly recommend incorporating the instance prompt with your prompt for better stylistic consistency with the dataset this model was trained on. So we should always include this stuff right here. So we simply copy and base this up in our prompting box. And now we can generate whatever we like, and my face should be on it, and now comes a really cool trick that I love. I simply go on Lexica and here I can search for stuff I like. I simply typed in man in a suit, and we can simply search for stuff that we like. Now, I assume I really like this picture right here. So I press on this picture, and we can simply copy the prompt, and then we can generate all of this ourself, and I hope that my face will be on it. So I go back into Leonardo so a character of Arnie, and then I include the prompt Italian gentlemen and tie compliment profile and so on. And I think we should totally just generate this and then see what we get out of this. And maybe I increase the output to four because you can see, we pay just one to generating these pictures. So we let run this a few times because the output, of course, is not always perfect. And of course, we can also include negative prompts if you like. So we include the negative prompts, and I include some stuff from my prompting list. And here we have some outputs. Now, I think some outputs are horrible and some outputs are somehow okay. You really need to play with this a lot. And here, I simply tried a few illustrations and so on, and I think the outputs are sometimes acceptable and sometimes they are completely awful. Yes, right now, stable difusion 1.5 is not a is completely perfect. But you can see that we can make pictures and the pictures look similar to me. Yes, the pictures they are a bit special to call it that way. But if you train the picture not on your face, but on other stuff, this will get better. Of course, I also didn't do a great job because I uploaded a lot of pictures with me in stuff like this. So I had like ten pictures, and I was always just sitting, and all of it was just basically screenshots from me right here on different backgrounds. If you want to do this really, really good, just upload a lot of good pictures of you. So in this video, you have learned how you can train your own models. You need to make pictures of you or the stuff that you want to train the model on. It's best to make roughly so 20 pictures, but the picture should be of every single angle, and it's also really, really cool if you use pictures where you're standing where you're sitting and much, much more and where the face is always of different angles. You should smile sometimes. You should smile a bit less sometimes. All of this is important. And I did not a great job providing a good data set because I was always sitting and just taking screenshots from me right here. Nearly every single time, we just had two pictures that were not screenshots for me just here, and then I simply changed the background. So I would really recommend you to do a better job than me, and then you will also get better outputs. So that's the training of a stable vision model inside of Leonardo AI. This is the easiest way. And you have also seen that, for example, the Spider model makes amazing pictures. And if you take good training data sets for your face, you can also make good pictures out of your face. Like I said right now, we train a stable division 1.5 model, and the 1.5 model is generally speaking, also not the model that makes absolutely the best outputs. So in this manner, just try to train your model. I think all of this is really, really cool and makes a lot of fun. And I hope also you have a lot of fun with these. And don't be sad if you can't make the perfect picture out of your face inside of Leonardo I because the training of the models inside of Leonardo is at least right now, not on the best level, but I would assume that this gets a lot better over time. So just try this out because I think this is cool. 103. Videos with Motion: In this video, I want to talk about Leonardo AI motion. We have the possibility to create our own videos inside of Leonardo AI. These are not exactly real videos, maybe not videos that you think of right now. These are simply four second long animations, but they look really really cool. Now I have to tell you right now. If you do this with faces, for example, it doesn't work that way. But if you use it with creative stuff, it's awesome. Just look with me at this. Now, this is some of the artwork that people make inside of Leonardo AI. You see some moves are really, really big, like in the left corner, and I think they are also a bit scary, and some moves are really sbtle and look enormous, good like this one in that picture. I think you need to really play with this a lot. Also, this ocean looks awesome to me. If you use faces, for example, like here, you need to add just subtle motions, or this will not work that great. So you really need to understand how the tool works. If you animated too strong, something like this will happen. Yes, it looks cool, but this thing completely collapses and loses all his original form. So how we can do this. This right here is just the community showcase and I pressed on motion. If you want to create yourself, you simply press on motion right here. Now you can either upload an image from your computer or for whatever you like. And then you can select from recent images if you like. If you have created something, you can simply select it right here, and then you can also add the motion strength, and you can make it public or not. Now, let's just do this with my little picture right here, and I show you something that we don't do. So normally, we don't use this picture and we increase it to maximum capacity. If we do this with something like humans, this is not great. And if we want to create this, it costs me 25 tokens right now. I simply press and generate and I think the output will not be good, so this is something that we probably shouldn't do right now. Creating pictures out of ourself is not the nicest thing to do, especially not with the motion strength at ten. And here are my generations. So you see, I got completely disfigured. And by the way, I simply made this twice. So I used the strength of ten with the same picture and everything, but the output is, of course, both of the times completely different. So this is totally random. You can't control this. You can just upload an image and press, create, and then basically you got what you got. This is Leonardo motion, and this is how we don't use it. And now I will show you how we can use it, for example. So we go back on Leonardo AI. I just go on all and now we search for something that I at least think would look cool if we animate it. Something like this can be really, really cool. So something creative. This works perfect. I simply can press image to motion right here, or I simply press remix, and then I can make the same picture once again here in my interface. So right now, I simply press generate and we will get exactly this picture or for example, four of these pictures. So let's just wait until we get our picture, and then I show you how we can animate every single picture also after we have created it in our normal interface. Now, here we get four pictures, and let's just see which of them we want to animate. Now, I think it would look really, really cool on this one, and we just go with the mouse over this tool. So you can see create motion video. So we don't use remove background, of course, we don't use image up scales, and we don't use alchemy up scales. You need to use generate motion video. We press on this, and then we can simply increase or decrease the strength just how we like it. I think we should use a strength of four example three. I have played with values of I think three to four gives us most of the time the best outputs. So the moves are gentle, and I think this looks really cool. We make this public and I press generate, and I think we should get a decent output with these. And in the meantime, while this is created, we can also search for other stuff. So you can simply go back on Leonardo AI. We can simply search for stuff here in the gallery, or of course, you can also press on motion. You can also search here for the stuff that you think is the coolest. Something like y fishes also work really, really nice in Leonardo motion. And here, we have a jellyfish. And because of this, I want to press on this jelly fish. And if you want to do motion to this jellyfish, you can either simply download it because motion is already added. Or you can simply recreate this. Also things like the ocean work really, really nice, and I think this should also work nice because in the background, we have some ocean. I just want to try this out. So I press on these and then we can simply use image to motion. So we don't make new pictures. We just press image to motion, and then we see what we get out of these. So everything is copied immediately in right here. Just remember as soon as we take pictures into account of people, use a smaller value right here, and then we press generate. Now, let's go back into my interface, so into image generation and see what we got previously because I think at least that we have our other output right now. Yes, and there we have it. And you see this looks awesome. I think you should do something like this with this Leonardo AI motion tool. So I think this looks really nice. We can also download it if we want. So we press download right here. And then we have this on our computer. And I also want to tell you a nice little trick. You can use this and throw it into our video editor, and you can simply play it backwards as soon as the video is finished, and it maybe looks something like it would be seamlessly and it would be run over and over again. Now let's just see what we get out of these. Yeah, it's somehow okay. So the hair wobbles a little bit. Also the water down here wobbles a little bit, but this thing on the leg is not perfect. So we don't have any control of these. This is the downside. In this video, you have learned how Leonardo AI motion works. It works really, really easy. If you have already created a picture, just simply press on motion, and then you get some motion. If you use low values, you get better outputs. If you use high values, the output will get somehow crazy, especially if you include people. If you include people and use a high value, the output will be completely awful. You should use stuff that is a bit more creative. Something also like the ocean or Jellyfishs work perfect. Play with values 2-6. But of course, there are no limits in stable division. You can also use persons and increase it to ten. If you think disfigured stuff is cool. So just try this out because we can make some nice little videos in Leonardo eye motion. 104. Conclusions of Stable Diffusion extras in leonardo: In this section, we took a look at all the cool features inside of stable difusion, and we used the Leonardo AI interface because this is at least in my mind, the best interface to use stable difusion. You saw the real time generations, the real time canvas, the normal Leonardo AI canvas, I painting, out painting, merging pictures together, image to image in the Canvas, sketch to image and how to add text. The three D texture generations for your three D printer, and of course, also how to train a stable division 1.5 model. All of this is really cool. And in every single thing, we use always different stable division models. For example, the real time generations, they work always with stable division X L turbo. This is a model that is enormous fast, and we trained our stable division 1.5 model. So my recommendation would be just try to use the stuff from this course. I think the editing in Leonardo is completely awesome. We dit pictures just like in photoshop, and this is really, really cheap to use. You have a lot of f generations every single day, at least right now as I am recording this course, and the subscription is really, really cheap. So I really think Leonardo is a nice tool, and you should try all of this out. And if you want to have some fun, just train your own model, train your stable revision, 1.5 model, because also this makes a lot of fun. And remember, if you want to make a model of your face, just take better training pictures that it because our output was not optimal. But also the stable revision 1.5 model makes not the best outputs. So have fun trying all of this out. I am 100% sure. You will have a lot of fun. This is your homework, and this is it for Leonardo AI. And if you have learned everything, you know what good learners do, they learn together. If you could tell somebody, Hey, Arnie has a cool course, that would be completely awesome. 105. What this section is all about, train your SDXL LoRA: In this section, we will learn how we can train our own stable difusion Laura. This is really, really cool, at least for me. We will train this inside Google Colab. All of this is completely for free. After we have trained our Laura, we can push it to hugging phase, and then we can download it also, we don't have necessarily to push it that hugging phase, and we can run our model at different interfaces. The easiest way is to let it run in C art. Stay tuned for the next lectures because I think this is awesome. You can train models on whatever you like. But I will make a model that is trained again on my face because I think we should totally try this out. That's basically also some of the hardest stuff that you can do. Training pictures on your own face is not that easy, but I think it's worth trying. And remember, the stuff that we are using right now is always the worst stuff that we ever get. All of this will get better and better over time, and of course, I will add new lectures in the sections as soon as this works even better. So stay tuned and see for yourself if you want to train your own stable division model. 106. Creating a Dataset to train your SDXL model: The first step is, of course, to make the right training data. If it's possible, it should be pictures about your face from different angles and in different environments. Also include some stuff where your whole body is seen, and you need to cut them in 512 512 pixels. Maybe you remember this web page. And basically, what we do right now is to upload our pictures and then simply cut them. I upload ten pictures right now, and I already have 20 other pictures, so I want to make this with a lot of different pictures, and we don't be serious. We just want to have some fun. I will simply cut these pictures a little bit, and I think we should get some nice little results. I have pictures from my body right now because in my last dataset I haven't included a lot of pictures with my body. So for that reason, I cut also some of these pictures where my body is included. And like I said, we just have some fun. It's nice to don't be serious with this stuff because come on, we only live once, so we include some really stupid pictures from me. I think we should have some fun with these. So these are simply some pictures. In some pictures, I don't have hair and nothing. In some pictures, I have a really, really stupid face. In some pictures, I am without clothes, so we simply want to download this. As soon as we have set what our frame should look like, we download this as a zip file. Then we have downloaded this, and now I will throw all of this into one single folder so that we can work really, really organized. So step number one was creating our data sets. It should be at least 20 pictures, but maybe better 30 or 40 pictures of our face in different environments, also maybe some different angles, some stupid faces, just to have some fun and also include maybe your upper body and in some pictures, also your legs. I have 20 pictures in the same resolutions of 512 by 512. I will add to this ten pictures, and then we have our data set. And I would recommend you to do the same thing. Now, you don't need to train a model on your face. You can train this model on whatever you like. Maybe you want to create a model that makes your cat or your dog. I think all of these are really, really cool use cases, and we should totally try this out. 107. Tipps on your Dataset: Real quick guy about the dataset that we created in the last video. It depends a bit on what cop notebook you work. Just make sure that every single picture has the same format. A B G format is perfect. Also, B and G's work in most of the stuff. You just need to look in the Cp notebook. Now, the outotran dream booth here, it's relatively simple and normally every single picture works. But just make sure that every single picture has the same format. This is really important. And if you are not sure, go into an online converting tool, type in into Google, convert in Cape B. And if you convert this stuff in CPAC and crop it in 512 by 512 pixels, you should run in no issues with all of these. So please make sure that you have the same format and ideally a Chap format. 108. Make a Huggingface token, API Key: In this video, we will create a token or an ABIKey inside of hugging face. I hope you remember what hugging phase is. Hugging face is simply the web page for developers, the community for developers, where we can train and save our own AI models. This is really, really nice. And if you don't have an account into hugging phase until now, it is really time to make one. So simply go on hugging face.com, and you will be on this web page. Now, this webpage you already know. I hope you remember the interface. It's really, really easy. Just make an account here. If you press on models, there are all the models that other people have trained, and of course, also I have trained a lot of models for hugging phase. Here, for example, you can see a few of my models, and we will train such a model. We will train a stable division el aura. If you simply press on settings, you can go on this right here, excess Docens. And here we can create a new doc. You simply press new doc. This right here is the API key for your hugging phase, and we need to give this a name. Let's just say best AI course. You need to change the role from read to write, and then you press generate a doke. Now your docen is created, and you can also copy your docen if you press on these. Right now, I will tell you, don't show this docen to other people, so later we will simply copy the doc. And that's basically it for this docen. So you need to make an account in hugging phase if you don't have an account right now. And then you simply create such a docun and you need to change read to write and don't share the docen with other people unless you want that everybody has access to your model. That's basically it, so do this right now. 109. Train your Stabe Diffusion XL Model (LoRA) with Dreambooth inside Google Colab Notebook: We have a lot of different Google Colab notebooks to make our training of our model. Right now, I will show you a quick overview of some of them, and then we will immediately train our model. This right here, this is the easiest cop notebook. It's called Auto train dream booth. You can see the cop notebook is really short. It's really straightforward and it works somewhat good. We will push our model to aging phase, and then we can run it. We have this STX L dream boo flora find tuner. This works really, really good. It's a bit more complicated. You need to include a few more things. But also this thing works really, really good. And the basic things are always completely the same. Here you can see also the output that this model can and will create. And then we also have the oia Laura dream booth trainer. Also, this trainer works really similar to all of this other stuff. I don't personally use this thing a lot because this is a lot bigger and we have to do a lot more stuff and don't really get much more better outputs. So we start with this trainer right here, out to train am booth because like I said, this is the easiest, and maybe we will also use this second one because the outputs get a little bit better. So let's just start with this one, like I said, this is the easiest one. And here you also have some steps that you need to do. So upload images to our folder named images. So that's the first thing that we will do. So we press on this thing right here. And first of all, we also should connect to our runtime. Now, you simply press on runtime runtime type, and now you need to use either the T four GPU or the A 100 GPU. It depends a bit on what model you use. So if you are in the free version, just use To gpu. And just because I am a good guy. I also use this right now. This takes a little bit longer than the paid subscription of the A 100 GPU. I just want to show you how long this takes if I use the free version. Now, I have this right here. I have the bigger rum. I will include this because it's a bit easier for me. So it's a bit faster in the course. If you don't have the subscription, maybe you can't use this right now, and now I want to save this. Now we will use our D four for this. Then we press this right here. And of course, we need to make a folder now. So we simply press new folder. We need to call this folder images. And here we need to upload our stuff. So the folder is called images, and this right here is my training data. I have 33 images. Every single one has a name, and I think this should work at least somewhat. I should get a nice output of these. I want to upload these images and we do this via drop. We simply copy all of these images and we drag and drop them into this folder right here. And you need to give permission. All of this will get deleted as soon as this Colab notebook is closed. Now I want to give this a name, and we simply delete this right here. Because the project name should be, for example, AI course, for example, then we need to use the model name. The SDX L is here perfect. Then we also need a prompt. Photo of a SKS dog, this is the standard prompt right here. You see, we should include names that are not typical that are not used on the Internet. For that reason, I use photo of a Haddle person. We should include if it's an animal, it's a dog, it's a person or whatever, and it's important that we use funny words right before this. I want to make all of this in small letters, so I think this should work. Photo of a Haddle person, this right here is the stuff that we need to include in our prompt. Now we should also push this to our hugging pace for that, we simply need to use here not false but true. The hugging face token here, we need to include our hugging face token. So we go back. We already made our doc, we simply copy this token. Then we go back into our hugging phase account, and we include our hugging face token. So we delete this and we put our token right here. Don't worry. I will delete the doc afterwards. Then we need to give our repo ID. So user name and repo name. And for that, we go simply here on hugging phase. So the user name is, of course, this right here. This is my user name. So I copy this, we go back into our notebook. The user name is this, and then we will create a repo name. The repo name can be whatever you like. Let's just say once again, AI course test, for example. And that's basically it. The hyperparameters, so the num steps we can and probably should also increase this a little bit. We can increase this up to 3,000. Yes, this is big and yes, this takes a lot of time. So you can also run this with 500, but you will get better outputs if you increase this just a little bit. As a rule of thumb, if you train on ten pictures, you can increase this to 100. If you train on 30 pictures just like me, you can also increase this two 3,000. For the C of this tutorial, I think I will simply use 2,500. Now, if you go down one more bit, see you can also run here an invariance interface. I don't think this works good. We will run this. Maybe in CRT, we will see where we run this because this right here doesn't work that great. Here, you also have some stuff that you can refine. But right now, none of this is needed. So the first thing is, of course, we press play here on the first cell, and we need to give permission even if the code comes from GTTub. Then you simply press play also on the second cell, and if you get an error on the first cell, just press play again. As soon as this is ready, so as soon as both of these cells are completely okay, we go down and now we press play on this outdoor trained rainbooth cell. And this right here, this will do all our heavy lifting. This will train our model, and this will take a lot of time. So you see all my images get loaded in right now. So we have my 33 images, I think, that all of them will get loaded into this outotrain booth, and then we will create everything completely automatically. Here you can see the models will get downloaded. So right now we download the model. And now the training starts. So we have right now 0%. So zero out of 2,500 steps. Yes, this will take some time. Right now I have one step out of 2,500, and this will take maybe one or 2 hours. Now, it's important if you are in the free version of Google Colab. It's possible that this Colab notebook gets shut down. And in order to don't let this shutdown, you need to do some stuff from time to time into this Colab notebook. So maybe stay on your PC. Maybe click right here and do something, delete something. You just need to be a little bit active from time to time in this notebook. Now, like I said, I have the paid subscription, so I can go away from the computer, and that's basically what I will do right now, and I will come back as soon as this is trained and then I will see you in the next video. And here we are again. I have to admit, yes, this took long time. But right now we have created 2,500 steps out of our 2,500 steps. And that's basically. So the whole training is done right now. And you have the possibility to test this out in this inferience interface, but I need to tell you right now, this inferience interface is awful. It doesn't really work, and because of that, I want to show you in the next video how we can run this thing. So in this video, you have learned how we can train our own stable diffusion, Laura. Everything we need to do is to go onto this nice little Colab notebook. We need to upload our own images, create a folder that is named images. Upload some pictures. In an ideal world, you should use a data set that is really, really awesome, and perfect. You should use pictures that are 512 512 pixels. You should have different faces. You should look somehow serious, somehow angry, somehow happy. You should simply have different facial expressions. You should be in different environments, in different cloth things and much much more. If you want to do this with your bet because I see a lot of stuff that people do with their bets. I think this works even better than with persons. So try to train a Laura with your bet and the same stuff applies. Just upload pictures. You need to use at least, but really at least five pictures, and you can go up to 40, and then you need to train the stuff, and the training is really easy. Just run all cells step by step and use a prompt. You need to give this basic prompt. So just use a word, for example, SKS dog as you saw in this nice little example. I used a different word. You just need to use a word that other people are not using, and then you will get good results. And in the next video, we will see how we can get our model and how we can test our model. 110. Download your SDXL LoRA Model from huggingface and run it in Seaart: In this video, I will show you where you find your model and how we can use it. So as you have shown, we have finished our complete training of our stable fusion, Laura, and this took a long time. Now we need to go on hugging phase. And on hugging phase, you need to refresh this page. So if you simply go on hugging phase, you need to go to models, and then you are in the wrong place. I would assume at least. Yes. Now here are the models of the other people. And I miss click right here and you need to watch this. So where you need to go. You need to go on your simple profile. You press on these, and here are all the models, you simply need to refresh this page. You see, this is our model, AI course test. I can press on this model, and now we can try to use this model also right here. You can type in some text and see if the model is working. Now I have to admit, most of the time the model is not working here. But sometimes it works. We simply try to type in our prompt here and see if we get something out. Now, you always need to use your starting prompt, and for your starting prompt, you simply press on files and versions to see what you need to use. Here you have the prompt text. You simply click on it. And here you can see what you need to use in your prompts. So you simply copy this, and then you can go on your model card and try to type in your prompt in this simple box. So try to type in this right here. Let's just say with sunglasses. Now, sometimes this works, sometimes it doesn't work, yes, we have a picture, but this is not our fine tuned model. We need to run this at other invariance API, you can simply do this by downloading your model. You need to download your model and then we will upload our model in a nice little interface. And that is really easy. You simply go on files and versions, and here you need to download your model. I would really recommend you to download every single thing because sometimes we need everything. It's also cool to have the read me, the prompts, and so on. So we simply download every single file that we have right here. And of course, the Laura is this right here. So the stuff that has the biggest size. So we download also the Laura and we simply doload everything. Now I need to show you another thing. If you go on settings, you can make this public or private. If you go on settings, this right here, make public. If you press this, every single person on hugging face will see your model. You model will be open source. People that go on this button that go on model can search and find your model and can download it. Now, if you don't want to have your model public, you need to make sure that this is private. Only you personal mode, or members of your organization can see and commit to this model. Right now, my model is private. If you want to use it, just hit me up and I will simply click Make Public. And as soon as I press make public, every single person can use my model, can download this model. This model right now is open source. I think I just make it private one more time because I don't know why people should make pictures out of me. But like I said, if you want to do this, just hit me up. So right now my model is private once again, and my model is also downloaded right now. So this gets now important because we go into CRT. Now, why we would go into CRT, because in CRT we can upload our models and we can use our models. And for that, we simply press here on train. So we can press on train, we can press on model, and we can press on Laura. And here we can upload and publish our models. If you press on train, you can also train your Laura right here. For that, you need to create a dataset. You can also make basically a similar thing right here in Card, but you need a subscription if you want to do this. So you can name. You can choose whatever you want to train, and then you need, of course, to create your dataset. But that's not what we do right now. Right now, we want to upload our model. And if we want to upload our model, we simply press on this plus sign right here, and we go to upload model. Now we need to name our model. Me in the AI, me in the AI, for example. Then we need to use a base model. We have trained the stable division X L model, so we need to chose this. Then we have trained a checkpoint, but a Laura. And now we need to upload our checkpoint file, and my file is right here in the downloads. I have created a folder. So we need to upload exactly this model, the Bida weight safe Densor. This is basically everything we need to do, so we simply press on these We search the Laura, we press on it, we open this up, and then the Laura will get uploaded right here. And normally this takes just like one or 2 minutes. I think this is right now completely uploaded, so we can press safe. And you see, as soon as we upload it, it's private. If you want to share this with the community, you can simply make it also public. Right now, I just want to save this. Save successfully, and then we simply try to run this thing. It's called me in the AI. Of course, we can publish it and make it published to all these people. But right now I don't want to try this. Right now, I simply want to press on this play button and test this out. We are right now in of course CRT here, what is epic realismus we can switch this, of course, and the Laura right now is in the AI. Right now, I just want to run one or two prompts and see what we get out of this. Remember, we always need to use this standard prompt to start, not with sunglasses but photo of a Haddle person. And then I simply add with sunglasses, we have a negative prompt, I simply let this run. Yes, this prompt is awful, but I just want to try this out. And here we get our pictures. And excuse me because I had a little error in the previous generation. Of course, here we need to use the right model. So we need to use stable difusion Excel. This right here is the Laura, me in the AI. We can increase and decrease the La strength. Now, I have played right now with a lot of different prompts and with a lot of different Laura weights and much, much more, and we got some outputs. Let me just just show you what we get. The first picture, I think this is really cool. Also, these two pictures, they are nearly perfect and look at this right here. These two pictures, at least in my mind, they are until now, at least the best pictures. Especially this in the left is maybe a little bit fat, but maybe that's just how I am. And the second picture, the picture in the right corner, I think this is like Nearly how I looked, let me know what do you think. Just take a look one more time. I think this diffusion model spotted the right stuff also here. So not all of the hair in the world. You can see everything right here. I think this pictures are nearly perfect. I like all of this four, and then we messed a little bit up with this two generations. Yeah, maybe they are somehow okay. Also this is somehow okay. So I tried to play a little bit. I tried to make some linking profile bed Some of them are somehow okay. Some of them are not really that good. I tried to increase and decrease the lower strength. I played with a little bit of stuff. Also this right here. Maybe maybe you can argue, yes, this is my face, but of course, completely different hairstyle and also, of course, completely different clothing and much much more. I simply tried to play a little bit. Now we have some really messed up stuff, or maybe that's just how I am. I think you need to tell me this. This right here was completely awful. I tried to increase the lower strength to the maximum capacity of two, and this was a complete mess. And then I simply tried to play a little bit. Some pictures are cool, some pictures are less cool. Then I tried to make an illustration. The first one was not that great. So yes, the output was good, but you can't see any of my face. Here, we get a little bit of better output. Maybe this right here looks somehow like me, and this is the last generation that I want to show you, and I think maybe also this looks somehow okay. Maybe not exactly my face, because the lower weight was a little bit too low, I think. So you simply need to try. You simply need to mess a bit. Just go scroll up at this picture because I think at least this picture was awesome. So this is basically just like me. You just need to play with the prompts, with the negative prompts, with the Laura strength, and much, much more. This is now just to try, but the training was nearly perfect. So in this video, you have learned how we can use our Laura. It's really, really easy. Yes, you can try it on hugging face, but most of the time, this is not working. The best and easiest way is to simply download your Laura. Upload it in Cart and here, you can run it, and you can run it completely for free every single day, at least for some generations. So you have a super cool interface inside of CRT Doti to run your nice little Laura. I think this is the way that we should use if we want to train a stable division Excel Laura. So just try this out. I think this is awesome. 111. More Colab Notebooks for LoRA training (a overview): So now you have learned how you can train your own stable difusion model or your stable difusion, Laura. We used this Colab notebook, so the outdo trained booth notebook. I would really recommend you to start with this notebook. This is the easiest notebook, and basically the output is the same as in every single other notebook. And yes, there are other notebooks. This is called the dream booth stable difusion. This notebook is completely the same stuff. You always need to press play on the first stuff. So first, check, of course, your GPU, then you need to run the install requirements. You give a token to your hugging face, and then you basically run this cell. Now, this notebook is especially for stable difusion di version 1.5. And it's possible that the outputs are not the dead good. Especially if you train this stuff on your face. Now, if you train it on apt, this may be works. You can also scroll down and see for yourself what you have right here. This basically is important. Here, you basically need to be feel prompt. So photo of a dog, of course, of a person, and the optimal cases, of course, also a small word this is unique for you. Then you press play on these basically, and you need to upload your images and then run this cell. So you can simply upload your data, and then you can run this cell. This cell is for training. Here you also need to change the prompt, so not a dog, but a person, and, of course, include your magic word, and everything else is fine right here, maybe except the steps. So the maximum train steps right here are 800. I think this is also enough, but of course you can increase it to like 2000. If you use 20 pictures, use 2000, and it's also possible to overtrain the Lura. If you use, for example, four or 5,000 steps, this brings not better results but worse results. Something 1000-3 thousand works most of the time. And of course, you need to render like forever. Then you can simply scroll down and the rest is basically done. You can specify the weights if you want, but you can also leave this blank. Then you can run this. Basically, you can test it, you simply press this cell right here and see if the output is okay. You can convert your weights, of course, to automatic 11 11. But of course, your model, you can also push it to hugging face and use it in C art just like you have learned. And you have also this inferios API right here. So you can also run this right here, if you press play on this button, and then here you basically type in your seat, You give the prompt, you give a negative prompt if you like. You give how many samples you would like to create the guidance scale, all of this, as you know it. And here you see, it's always the same dog, so I would guess that always the same person has created this nice little colab notebook. I also think he has nailed down his bed. Of course, maybe if you want a good model, you need to run this a few times. And then you have also this here. So the SDX L m both Laura and S every single time. This works just like the other cop notebooks. So press play on this first button, you need to give permission. You need to choose the right run time. You need to use D four or maybe if you have a bigger graphic card, use also a stronger one, then you scroll down, you play here. On the install requirements. You import everything that you need, then you need to upload your dataset right here. He has also told you some lines, what you need to do. So all of this works completely the same. And if you want to use images from hugging face, you can also do it this way. But most of the time you will basically upload something like faces from you or from your bed. That's at least what I see a lot of people do. You can also train a Laura on your wife, on your man, on whatever you like from the bird that you see outside if you have a good data set, Then you can press to preview images, and here you can see the preview images off this nice little pet. So he has a good dataset, and he creates nice and standing outputs with these. And as soon as everything is tested, you can simply start with the next cell. So like I told you, this notebook is a little bit bigger, but we just need to change the prompt. We change the file of your images, and then you are basically also ready to go. So we scroll down right here. Remember, always press play until you come here to train, and then you can simply train also right here. You nice little stable diffusion X L, Laura. As you see it right here, so we train on stable difusion X L. And of course here, you need to name for the first of all your dataset. You need to give a instance prompt, so of course, not the dog, but a person or whatever you want to train, and you cannot play a little bit with the steps and the steps are right here. So every single time, the same stuff. And as soon as this is done, you can also save your model. You can either push it to hugging phase, always the same thing, so push it to hugging phase. You need to give your hugging phase token, and that's basically it. And then you can also run it on this inferience interface. But most of the time, I would not recommend you to do this. Yes, you can test it, but I think it's better to run it in an interface like CRT. So this video was basically just to show you that we have more options. Now, if you really think you need a p dive in this cop notebooks, just tell me and I will include them. But like I said, they work completely the same, and you get also basically like the same model if you use the same dataset. I would really recommend you to use the easiest thing right here. That's because why we started with the easiest thing. I see a lot of videos out there, and they always try to use the most complex cop notebook. No. Complexity brings you not better outputs. You get better outputs by better data set. So make sure to prepare a really, really good data set. Don't over train your model, but also don't undertrain your model, use something like 1,000 steps or maybe 2000 2,500 from our video box. Also, maybe a little bit of an overkill. You need to find your sweet spot. I would assume that you need to train three or four auras. You need to take some time, and then you have your perfect model. Then you have your stable difusion Excel Laura, and you can run it wherever you like. You can use such a inferien interface, but I think they are awful just run it in Z art and have some fun with your new stuff. 112. Conclusions of Fine Tuning a Stable Diffusion Extra Large LoRA with dreambooth colab notebook: Over the last lectures, you have learned really really cool stuff, at least in my mind. You have learned how to train a stable division XL model. This stable division el model is not a really a model, it's a Laura, but you can use this Laura and throw it over a normal stable division XL model, and you can create whatever you like. Pictures of you, pictures of people you like, pictures even of celebrities. You can train such a model on celebrities and make, for example, Donald Trump. I don't think if this is the nicest thing to do, but of course you can totally do this, and you can nail this things down if you do it the right way. First, you have created a dataset. Different environments, different facial expressions, and of course, you need to crop this stuff in 512 by 512 pixels. Make sure you use B as a format. This works the best. Then you made the hugging phase account and you generated a hugging phase token because we need this token in nearly every single colab notebook in that we work. And then we basically did the heavy lifting with dream booth Auto training because this is the easiest thing. But I also gave you the quick overlook to all the other cop notebooks that are out there to fine tune your stable division Excel Laura. I think I have a cool homework for you. Just try this stuff yourself out. Just go make some pictures of you or from whoever you like, maybe from your pet and train your own stable fusion Excel, Laura. I think this is, like, really, really fun, and you should totally try this out. And I have one little question for you. It would mean the world to me if you can share or rate this course. This would be enormous nice for me and for you because also other people will get value out of you if you share this. So thank you for your commitment.