How to use DALL-E 2 to edit and create images | Jonathon Parker | Skillshare
Drawer
Search

Playback Speed


  • 0.5x
  • 1x (Normal)
  • 1.25x
  • 1.5x
  • 2x

How to use DALL-E 2 to edit and create images

teacher avatar Jonathon Parker, Passionate MoGraph and VFX Lectu

Watch this class and thousands more

Get unlimited access to every class
Taught by industry leaders & working professionals
Topics include illustration, design, photography, and more

Watch this class and thousands more

Get unlimited access to every class
Taught by industry leaders & working professionals
Topics include illustration, design, photography, and more

Lessons in This Class

    • 1.

      Course Promo

      1:49

    • 2.

      What is DALL-E 2

      4:09

    • 3.

      How to access DALL-E 2

      1:51

    • 4.

      How to generate images in DALL-E 2

      10:47

    • 5.

      Using DALL-E 2 with GFPGAN to create realistic faces

      6:04

    • 6.

      Prompt Tips

      6:35

    • 7.

      First Class Project

      1:13

    • 8.

      Creating Variations from images in DALL-E 2

      6:30

    • 9.

      How to replace part of an image in DALL-E 2

      7:24

    • 10.

      Changing the environment of an image in DALL-E 2

      7:41

    • 11.

      How to add a person to a photo in DALL-E 2

      10:26

    • 12.

      Expanding the edges of an image in DALL-E 2

      7:28

    • 13.

      Final Class Project

      1:09

  • --
  • Beginner level
  • Intermediate level
  • Advanced level
  • All levels

Community Generated

The level is determined by a majority opinion of students who have reviewed this class. The teacher's recommendation is shown until at least 5 student responses are collected.

291

Students

4

Projects

About This Class

DALL-E 2 is one of the most powerful AI Art Generation programs on the market. It is comparable to that of Midjourney and Disco Diffusion, however DALL-E 2 has tools to not only create generative art, but to edit existing images. The ability to edit existing images is what sets DALL-E 2 apart from the crowd. 

This course will be split up into two sections, firstly the course aims to simplify the process of utilising text prompts along with additional parameters to create stunning pieces of art. Secondly this course will demonstrate the numerous ways in which DALL-E 2 can be used as an image editor, by removing or adding things to existing images (known as inpainting). We will also discover how to extend images beyond their original borders (known as outpainting). 

DALL-E 2 has now removed their Waitlist and the software is free of charge, There is however a 15 credit limit per month. Within this course I have purchased extra credits as I use DALL-E 2 a lot in my professional work. However there is no requirement for you to make any purchases!

The only pre-requisite to this course is that you need an email account and access to the internet. I do also use photoshop in one of the lessons, but again this is not a requirement for the course.

Course outline

Part 1 Using DALL-E 2 to generate AI images

What is DALL-E2? (A walkthrough and introduction to DALL-E 2)

How to access DALL-E 2 (This session will show you how to access DALL-E 2)

How to generate images in DALL-E 2 (In this session we will utilise the prompt command to generate our first artwork)

Generating faces in DALL-E 2 and how to use GFPGAN to correct those faces (In this session we will utilise an additional AI tool to correct some of the wacky faces DALL-E 2 generates)

Prompt Tips (This session will cover some additional tips for generating text prompts)

First Class Project – (This lesson will walk you through the first assignment that you have been tasked with, summarising the skills and techniques learnt so far in the course.)



Part 2 Using DALL-E 2 to edit images


Creating Variations from images in DALL-E 2
- (In this lesson we will upload existing images to see what variations DALL-E 2 can make of the image)

How to replace part of an image in DALL-E 2 - (In this lesson we will use the edit function of DALL-E 2 to replace specific parts of an image)

Changing the environment of an image in DALL-E 2 - (In this lesson we will use the edit function to change the environment of an existing image)

How to add a person to a photo in DALL-E 2 - (In this lesson we will use DALL-E 2 and GFPGAN to add a person to an existing image)

Expanding the edges of an image in DALL-E 2 - (In this lesson we will use the generation frame function to extend the edges of an existing image)

Final Class Project(This lesson will walk you through the final assignment that you have been tasked with, summarising the skills and techniques learnt previously in the course.)


As mentioned above in the course outline, the course will finish with a really fun assignment where you will have the opportunity to practice the skills that you have learnt in this course! I'm looking forward to seeing what you guys create so please upload your work!

I hope you enjoy this course as much as I enjoyed creating it!




For the Class projects, please see a detailed outline in 'Project Task (Assignment)'.

Meet Your Teacher

Teacher Profile Image

Jonathon Parker

Passionate MoGraph and VFX Lectu

Teacher

My name is Jonathon Parker, a Motion Graphics and Visual Effects lecturer from the UK.

 

Firstly, a little background about me! I have always been the creative type and always strive to improve and push my creative skills further and further.

 

As stated above I am already teaching in the UK as a Motion Graphics and VFX lecturer – The difference with my courses will be the fact that they come from an educator, not someone doing this just in their spare time! All my courses focus on teaching and assessment. I will teach you a few skills and then set you an assignment to check that those skills have been learnt. I also see the importance of generating handouts and resources for the student, so I have included these in my courses also!

&nb... See full profile

Level: Beginner

Class Ratings

Expectations Met?
    Exceeded!
  • 0%
  • Yes
  • 0%
  • Somewhat
  • 0%
  • Not really
  • 0%

Why Join Skillshare?

Take award-winning Skillshare Original Classes

Each class has short lessons, hands-on projects

Your membership supports Skillshare teachers

Learn From Anywhere

Take classes on the go with the Skillshare app. Stream or download to watch on the plane, the subway, or wherever you learn best.

Transcripts

1. Course Promo: Hi guys and welcome to this course on Darley two. So what is Darley to? Well, w2 is an AI or more specifically, it's a neural network that takes text prompts and generates images from those texts bonds. But W2 is a bit bigger and better than stuff like mid-June and some of the other ones like that. In my opinion, the only two has extra options where you can edit parts of images, edit parts of images out, and do a lot more than just generate images from scratch from textbooks. Now, Darley to as very recently gotten rid of the waitlist, which means anyone can access and use daily to all you need is an internet connection and an email to sign up with. This course will cover a wide range of things. We'll start off by looking at how to generate images from textbooks. Then we'll start to look at how we manage bases within AI. Because what Darley to and other AI I'm good at yet is generating realistic basis. So have a look at how we manage and control that. We'll look at Tips for Generating prompts. We'll look at how to remove things from photos, how to place things in new environments. How to add people, hold new people into photos, how to extend the edges of photos. And amongst all of that, there will also be class projects where you guys can generate stuff and upload it for feedback from me. And you can also look at what everyone else has been doing. So I'm really excited about this course, guys. I've done other AI courses, but haven't done one on Darley yet. And this is gonna be the first one. I'm really looking forward to seeing what you guys think and what you guys do with this course. So there's nothing left to do but jump on in. 2. What is DALL-E 2: Okay, So welcome to the first class in this course. Now what we're going to look at in this class just very quickly is what is Dolly and actually how does it work? So if we go to openai.com and go to the blog, we can see the Darley, this is their website, is a neural network that creates images from text captions from over a wide range of concepts. So essentially it's converting text prompts to images. However, Dolly actually does a lot more than that. Now, just thinking about the name. It's actually named after Salvador Dali. So that's just a little bit of extra interesting information for you there. But yes, like I said, it uses image. It creates images from text. But Darley, like I said, doesn't just create images. It also edit images. So e.g. if I go to here, you can see this is an image which has been uploaded and actually you can erase part of an image and start adding to that image and generating more stuff in that area. So e.g. you could have a photograph of somebody with glasses. He raised the glasses and just replace that just with a normal human face. Not that there's anything not normal about glasses as I wear them myself. Now, this is Darley, when you login to it, what it looks like and you can start typing things away in here and either generate an image. And like I said, Oh, you can upload an image to edit. If we hover over this, it says an armchair and the shape of an avocado, that would be a text prompt which somebody typed in. And it generated this an expressive oil painting of a basketball player Dunkin depicted as an explosion of a nebula wireless quite in depth. And that's what happened there. And we've got more examples. Here. Is article by here, which I'll link you to. These incredible images were made with Darley to these two images here. This is not a photograph. So let's start off. Dog doesn't exist. Portrait of a Bassett hound, eight K photograph. That was the request which someone typed in, and this is the image that got another one. Here's another image where Dolly was asked to create a photo, a tasty VLT. We've got another one, diorama of a ship in a bottle detailed and Bokeh, bokeh be in these out-of-focus elements in the background. And again, a floating chair. Just look at this one. The shadows that daily is attempted to recreate, maybe not perfect though shadows, but just the idea that you've got something floating and it knows that there should be shadows there. Even though it's floating in the air. Leonardo enters the metal verse that's quite an interesting one. Butterfly perched on the central slip and dragging digital art. I mean, look how awesome that is. So again, I could kind of go through these forever, but not all. So is it, like I said, is it able to be used for just generated images from scratch? But you can, like we said, upload an image to edit it and erase certain parts of those images. So if e.g. you will stand in a photograph and there was a blank background behind you of the sea and the water. Perhaps you could then erase that and type in man standing by pyramids. And then it just, it just added the permits and the background of that image. Now this course is going to be looking for, first of all, how to generate images from scratch using prompt. But later on we're going to look at replacing parts of an image, removing objects and images, removing a person from a photo, added a person to a photo and all this kind of stuff. Expanding, expanding the ages of photos so that if you've got a photo of somebody and you just want to make that photo bigger. Well, you can add stuff to the edges, just like we're doing here. And it's called Generation frames. So there's loads that you can do with Darley and it's really, really incredible tool. The next session is going to go through how to access Dolly because it is completely free. There are payments solutions where you can get more credit, but I'm going to be going through that in the next lesson. Chase for joining, and I'll see you in that lesson. 3. How to access DALL-E 2: Okay, So in terms of access in Dali, what you're going to want to do is you're going to want to navigate to this web link, openai.com forward slash double dash, E, dash two. Now I'm going to put a text document. I'm going to put a text document with the link for this, but if you want, you can just type. Click here. Now I would normally go to login, but you guys, if you haven't used as bore going to want to go to sign up. Once you get to sign up, you can log in with a Google account or Microsoft account or just kind of make you e-mail address and so on, so forth. I'm not going to walk you through that bit, but Dolly is now open to all members. Now I'm going to go ahead and login. So now that I'm logged in, you can actually just go away and start creating and start uploading images to Edit. And I'll show you a lot of stuff. But one thing that you're going to want to be aware of, in terms of a free account, what you get is you get a number of credits and each time you click the Generate button when we're editing or generating images, you use one credit. Now you'll notice that if I come here, I've got 99 credits. Okay. Because I've purchased I think it was 115 credits for $15. Because 15 credits the free ones which you get monthly, they'll go quite fast. But there's no need. You don't have to purchase elements, you don't have to pay. But if you want to do more and more which, which I tend to do a lot of, then you may end up purchasing some, but it's absolutely an option. You can use this free of charge, but there is a limit. So that's going to leave it there, to be honest, because in the next session I'm then going to go through using prompt to actually start generating images. So like I said, and leave it here and I'll see you in the next lesson. 4. How to generate images in DALL-E 2: Okay, So we're back into Dolly. And what we're gonna be doing in this lesson is we're gonna be looking at using just image prompts to generate artwork with inside of Darley. And we'll have a look at some of the additional features and some are perhaps the weaknesses. So I'm going to start off by typing in like, what do I want to see? Actually, you'll notice that if we just have a look at this e.g. it's saying an impressionist oil painting of sunflowers in a purple bars. So it's describing some stuff over here, but it's giving it a style as well. If we look here, 3D render, That's a style of acute tropical depression and aquarium. So it's the same what it wants to see. And then he's got a style at the end. So you'll quite often see descriptions of what we want to see, but also the style we want to see it in a blue orange slice. That's not given style. This one is depicted as expressive oil painted. Some of them do, some of them don't. So if you're not happy with your results, you can try and add a few different things. So I'm going to go ahead and type a. Larvae. Can't remember how to spell correctly. I am going to go ahead and double are driving down a desert road. And once, once you're happy with that, you can insert commas and then see if any of these columns yet insert commas as little breaks and then look at the first phrase and then look at the second phrase separate. But I'm just gonna go for this and I'm going to click Generate. I think that's how you spell for ROE. If not, then maybe I won't get too far away. I'm sure I will. Another interesting thing, it does give you these tips. So ask for mediums like oil pastel pencil. Okay, Now this is pretty cool. We can click them to view them individually. You can tell us a little bit of Bacon's good on Next, it looks like this is a photograph, but this is almost like digital art. And again, this is pretty cool, but it's not photo realistic. So I'm gonna go ahead and add a comma photo realistic. And I'm going to say a photograph of a Ferrari driving down the desert road. Photorealistic. I haven't tried this, but I'm going to see kind of what these results get me. Because again, if I'm not happy, what are sometimes try and do is be more specific with the actual texts that I'm giving it. Now. This one isn't a photograph. I mean, there's photograph photographic stuff in there. But I don't mind this one. So what I may do with this one, well, actually what I can do is I'm gonna go and just type in eight K and detailed at the end. Just to, just to try something else, I'm going to click Generate again. Let's just check that that's been spelled. There we go. It's given capital F. Interesting that it didn't give me actual photographs back then. Now, here we go. We've got some stuff which I kind of do like this one. It's more photographic but still looks a little still. The detail here isn't quite there, but we're getting there. This one I don't like this one. I do like as well. It does still look a little bit digital, but you know why it's not terrible. And this one again, it still looks a bit digital. So I think I'm gonna pick either this one or this one. Let's go with, let's go with this one because I like the blue sky in there. Now what I can do is I can click, Okay, Let's do some variations. And what it'll do is it'll take that one and it'll use the same prompt and it'll generate very similar variations based on that image. So here we've got our variations based on that image. You can see obviously that they're quite similar. Now, what you can do is you can generate variations of a specific one. You can download them, you can edit that image. Now we're going to have a look at editing the image images later on. But again, you could say right now I want more variations of this. You could go ahead and just save that one out. But this is the kind of thing that we're dealing with. I don't like these two. Maybe I like the composition of this one a bit better. So I'd kinda go ahead and generate variations of this one. And you'll use AP credits quite fast, but this is quite a powerful or while ago already seen piece of kit. So what I'm gonna do is I'm gonna go ahead and I'm going to try another one. I'm going to go a VHS player in the style of vapor wave. Now vapor wave is an art style. And I'm just going to go ahead and see what this gives me what I'm doing differently this time, I'm starting off with a art style. Okay, So this is what we've got. I don't really know why there's faces in these two, but this is the kind of thing I was after, looks very 80s, 90s. So I would probably go ahead and click this one and generate some variations based on this one. Also note how well it's doing the reflections, but here that is pretty cool. It knows that this split is a darker shade, so should be darker reflection. And notice this is lighter. It's capturing very similar stuff here, not perfect, but very, very similar. So I'm gonna go ahead and generate some variations of this now to see what else we can get. Okay, So this is definitely more like this is what I was going for. I actually think so far I prefer this one or this one, but this is definitely obviously giving you a very specific style. So let's try something different. Let's go for another photograph and I'm going to try and show you what happens with faces sometimes because one thing it's not very good at is faces. However, in a future lesson in this course, I think it's actually gonna be the next lesson. Yes, the next lesson I'm going to show you how to actually deal with that. So let's go ahead and say photograph of the mad scientist. Photograph of a mad scientist. And I'm gonna go ahead and say, just put eight K in there so it knows I want to detail, Let's click Generate. Okay, So these are the results. Now I'm going to show you why dolly isn't great at faces yet. So if we click this one, looks pretty cool. You have a look at his eye. His eyes. Dolly is not greater than ice. Okay. It looks very uncanny, like uncanny valley. If we look at this one, again, if you just cover up his eyes for this part down it looks, it looks pretty normal and it does look like a mad scientist, especially with those eyes. So again, eyes and quite often teeth dolly is not very good at. And I'm gonna give you another example of this. So if I type in a cute couple smiling by the seaside, and click Generate to have a look at what monstrosity it might deliver. Okay, So this is the results. And again, you can see, I just can't do faces very well. Let's have another look at this one. Yeah, the eyes just aren't great. Teeth aren't too bad. This one, again, there's a few ways we could actually solve this. Let's say e.g. you end up really liking this photo background and everything, or this one perhaps it's just the eyes and everything. Well, we could go ahead and solve that. And that's something which we will be doing like I said in the next lesson. Let's have a look sweet to us. Do another one more 11 prompts that I often like doing is to test new AIs. Is cyber punk, city, metro talking about in the future. I like and I really like an artist called Simon style and hug by Simon style and bog. And I'm also going to put trending on art station from 0. This particular prompt trended on art station. What the AI would end up doing is looking at the kind of artwork that trends on art station. Okay, now there's loads of prompts, tips and again, that'll be in two lessons time when I give you prompts, tips and there's loads that you could look at. But yeah, this is what I always like to give ai just, just as a bit of a baseline. If I give this prompts to every A-I, I can see the differences between each first going to have one final look at what this gives us. Okay, So this is pretty cool. I am absolutely loving. I'm loving all of them. This one is really, really cool. I'm loving these ones over here as well. Now, I'm gonna give you a little bit extra, let's say just, just for argument's sake that we don't like what's going on in I don't know. Let's go for this one. Okay. And let's say we want to erase these buildings, but what we can do is we can edit this photo, get our brush, and start painting out the bits which we don't like. Okay, this is something I'm gonna go a lot more into with actual real photographs. And let's just erase this stuff up here and say, Look, I liked the foreground, but don't like what's going on in the background. What we can do is just go like that. And then if we keep the same, you can change this up here. But if we just keep the same and click Generate, what is actually going to do is it's now going to use that same prompt and generate the missing part of that photo. This is something like I said, we're gonna go a lot more depth into, but I could have changed the prompt and I could have gone from Van Gogh's Starry Night and they would have done something like that up there. So again, this is the original. You can see now it's kept the bottom and it's actually are, I'm loving this, I am loving this, but here is actually generated different backgrounds based on them. Okay, so that's just a little bit extra. Now, go ahead and play with this to your heart's content. The next lesson is going to be about generating images with faces and an additional AI we can use to correct those faces. And then there will be less than on prompt tips. And how to get better prompts and use different keywords to generate better prompts. And then there will be a little bit of a class assignment before a load of other stuff. And that's just part one of the course. So yeah, hope you enjoyed this. I'm just on the right butt here. By the way, you've got a history Barb, all the things you've ever generated. So you can see, I went the other day and I was saying a brown teddy bear in a town center and these look like photos, like literally that looks like a photograph. But it's not that that was never a photograph. Okay. So anyway, I could go on for days, but yeah, I look forward to seeing you in the next lesson. 5. Using DALL-E 2 with GFPGAN to create realistic faces: Hi folks, and welcome to the lesson on generating images with faces and using an additional AI to correct those faces. We learned in the last lesson. The AI isn't great with paces, didn't we? So I'm gonna go ahead and type in a prompt. I'm going to go man standing beside peer or man standing beside beach, smiling camera. Okay, so hopefully we should get a guy who smiled and maybe showing his teeth, which will be good because that's an area that AI in great with. So let's go ahead and let that generate. Okay, So this is what we've got. Pretty hilarious if I'm honest. Click-through. Okay, so we've got some stuff there. I may work with this one to be honest. We've got some interesting stuff there and you can tell that it's just the face. I mean, don't know what's going on there. I may even use this one. Now what you can do is you can go ahead. And what I would always say is try and get Darley to get it to get it as good as possible. What you could do is you could go ahead and say, Okay, I'm going to edit this. Let's maybe lower our brush size. I like this photo, but I just don't like these eyes and these teeth. So I'm gonna go ahead and keep this up here, that, that prompt in there and click Generate. And what it's gonna do is it's going to take that original and try and generate some variations of those teeth and the eyes. And if we can get it to the best possible starting point, then the additional AI software that we use won't have to do as much work. Okay, so we've got some options. We've got this, which isn't too terrible to be honest, that this and this, I'll be honest, they're not terrible. So what am I do is I might actually take it, may take the original. I'm going to download the original just to give it a little bit of a worst one because there's a lot that needs to be solved there. So what I'm gonna do is I'm going to click the original. I'm going to click these things and I'm going to go, Oh, I should be able to download that one. What I might want to do is I might want to pop back in my history tab and download it from here. There we go. I've now got that downloaded and hopefully somewhere there we go, Dolly. And what I'm gonna do is I'm very quickly going to drag that to my desktop and I'm going to name it breaks correction. So what I wanna do now is I'm going to navigate to another website. If you were to type, just go to Google GFP, Gam and type that in. And just this link over here, GFP Gan, aimed at developing practical algorithms is a GitHub link. Now what you are going to need is you're going to need a GitHub login. But if you haven't done it, don't do it yet. You want to scroll down here and click updated online demo. So just click this demo and it's going to open a new window. Now this is completely free as I'm recording this on 9 October 2022. So this is completely free and you can see kind of what it's doing here. Now you may need to log in at this point to do it. And as soon as you've made a login and you've got this link, what you'll wanna do is it says drop or click to select a file. So I'll click here and I'll just go desktop. And where is it face correction. And I'm going to click Open. I'm going to open this image once I've logged in and everything. And then I'm going to say, yeah, these affine scaling factor and that version, think that's most of the most recent version yet. I'm just going to click submit. And what it's gonna do is it's going to do its work and we'll see just how close it gets. It all. Have a look at the potential improvements and I'm hoping that it has improved a bit. I mean, look at that to that, that is a lot better. In fact, that, that's actually quite crazy. Look at this eye compared to this side. What I can do now is I can actually download this input. One of the things that this software does do is it does sort of smooth a lot a bit as well. So the skin here is looking a lot smoother than over here. So if you had some sort of Adobe Photoshop or GIMP or any photo editor, you could place this image on top of this one and just mask in the face areas. But again, what an incredible improvement. This is just another sort of AI tool that's out there and I'm going to download that because that is a very, very good alteration and just look at the teeth as well at teeth are ten times better. Now imagine if we were back in here and we'd have used, I don't know, perhaps this, maybe it'll hopefully brought the eyes down a little bit more. Let's go ahead and put that to the test as we're here. So I'm going to download this. I'm going to come back into here is called. Let's just drag this to our desktop again. And what we're going to call it base correction. To misspell that. Oh, well, I'm going to refresh this page. I'm going to go face correction to. And I'm going to click Submit and then see what happens this time. Because we use Darley to try and correct it the first time. And I wonder maybe using the original, maybe better. And again, this is a little bit better. They've made the ice smaller. But I actually prefer when we use the original version. Again, you guys have got options. So that's it for this lesson. In the next lesson, I'm going to give you some prompts, tips, some keywords you can use, and some good different art styles you can consider using. After that, I'm gonna give you a class task, and then we're going to jump into part two of this course where we look at a lot more in detail at the edit functions of Darley and how we can really, really utilize those functions. Okay, Cheers guys, and I look forward to seeing you in the next lesson. 6. Prompt Tips: Okay guys, so in this lesson, I'm going to give you just a few of my own personal pump tips and a big sort of really handy resource at the end, like a big PDF, which will help you a lot. So from my experience, some of the words that I use because I like to generate a little fantasy and sci-fi stuff. If I use the keywords station and just add station. So this tends to give me some sort of fantasy, fantasy inside by vibes. Now, if I take you to art station, Let's just go ahead and go to our station. Soften up the word trend in on art station. This is the kind of artwork which it will be looking at. So again, trend in on art station, there'll be looking at all this kind of stuff by here. So that's my first tip. Another tip is to use artist's names. So again, we proved a few of us may have our own preferred artists. So I like the artist Simon style and hog, especially for sort of environment stuff. And if you were to have a look at his work, you probably won't be surprised why I love his stuff is just so cinematic and just yet amazing. I mean, look at, look at this stuff, so cool. So again, you may have some artists that you really like. I don't know, Maybe Andy Warhol, e.g. if you were to type that in as part of the prompt, I think you'd start getting some stuff that looks a lot like this. Okay? Another thing I would do is try short versus long prompts. So you could try. I don't know if you put man standing at gate that could give you like just a bloke standing at gates. But you may want to go for something Gothic. So you could say, man, tall, dark skinny man in a cloak in the shadows, standing by a rusty falling apart gate in a dark, gloomy environment, that whole thing would give it a lot more. Sometimes short prompts work really, really nice if they're really, really vague. If you were to type in the wind, I have absolutely no idea what Darley would give you. But if you were to type in windy day on a desert island plot palm trees bending in the wind, debris flying everywhere. Obviously that's more specific. So I sometimes like short prompts, but I also sometimes like long prompt as well. So try using short versus long prompt and see what difference you get. Another thing that I'd like to do is time of day. If I want something to look quite cinematic, I know as somebody who teaches film that a lot of Netflix stuff is shot in the golden hour, whether that's sunrise or sunset or just before sunrise and just before and just after sunset. Because you get those sort of hazy, hazy looks. So if you're doing some environments, stuff you may want to do at sunset or early in the morning and give it a time of day. Another one, art movements, slash styles. And I think that's a big one. So if e.g. we were to go and have a look at top ten movements. Talk to an art movements. And let's go ahead and find an article. Top ten most influential art movements. And it'll probably say, look, here's line art again, that gives you a specific style. Then what else? Impressionism, post-impressionism. You may just want to generate photos. But pointillism. Let's have a look on the VO, Cubism and all this kinda stuff. So you may want to go for an art style, top ten digital art movements. I wonder what that is. 17 digital artists you need to know about. Because again, you may find an artist's style like Mandy Joggins. You may click this link and you may love her work so you can use her name and your prompts. Or Irene a French have Aldo Ortega, Rachel Walpole. You may like, I don't know. You may like a certain computer game. Like, I know I love art world. This whole franchise, I love this slide may type in these as little things. I may find these artists names and type them in as prompts. Okay, there's loads of stuff to think about there. Now the last one is a little link which I'm going to give you as part of this lesson. And if you type in this link, It's a PDF. And basically here is a 82 page prompt book. So Scott, you can scroll through and there's some stuff in here to read. This contains photos that aren't real photos, but obviously you've got stuff in here so you can think right, okay, Let's go photography. Let's have a look at what words can be used to implement that. So let's just get photography. And it'll say extreme close-up, close-up. All of these stuff gives you different words. Camera angles, you know what you may think are aerial view of something that would be cool. You could type in the word Boca and that'll give you this nice out with depth, depth of field that will focus lock, motion, blur, tilt chip photography. This may give you, this may give you, excuse me, loads of things to think about. Flash photography, colorful lighting, and width that with each term, There's additional words. So if you want it cold, something that's lit quite coldly, you could put fluorescent light in 4,800 K, all that kind of stuff. So let's have a look at general vibes. You, these are all emotional prompt to it. So for positive mood and low-energy, you could type in all of these ethereal, elegant, soft light. If you want something positive, you go for all these words. There's loads of stuff in here. Let's have a look at illustration, what we got here. So stencil art, these are all different styles. Pencil sketch, pencil, sketch versus drawing, different things. I wouldn't realize that. Again, loads of stuff in here. So crayon, you could type in all of these steps and this is gonna be really handy resource and I will upload the link for you for this. This is made by Gallery, Gallery. So it's just a website here. So you can have a look through that, all that other stuff that here's that prompt book. But yeah, this is probably going to be a really big resource for you in terms of generating prompts. So I really do hope that that gives you some help. Tune into the next lesson where I'll be sending you the first class assignment for this course. So choose for tuning in, and I'll see you in the next lesson. 7. First Class Project: Hi guys. So for your first class project, I want two things from you uploaded two images using the text to image prompts generator, which we've been going over what you're typing a little bit texts and you've got some images. I want an environment and a character uploaded from all of you, along with a little bit of text terminals, what prompts that you use because then we can almost use that as a bit of a gallery and see what prompts are working for other people. It becomes a little bit of a resource hub. So like I said, two images with the prompts that you use. The first image, I want it to be an environment. It can be a photo realistic environment. It can be abstract fantasy in any art style, whatever you want. The second image, I want it to be some sort of character that can be a person, an animal, an alien, photorealistic, imaginary, abstract, any art style, whatever you want. Okay, so that's all you gotta do for the first project. Two images tell us the text prompts that you use, but I want one environment on one character. I hope you guys enjoy that and I can't wait to see what you guys make. 8. Creating Variations from images in DALL-E 2: Okay, so welcome back to part two of this course where we're going to focus a lot more on the edit functions within Dolly, I hope you enjoyed the class task and I hope you managed to create some really cool piece of art work. But in this section of the course, we're going to be really dive in into the Edit function, which I think is probably the more powerful part of Darley, if I'm honest, and what makes it stand apart from other ai are two generators out there. So I've downloaded a bunch of images to work with for this lesson. I've downloaded an image of a ship and image of mountains and image of a football dog and a cat and squirrel. And so you could have used whatever images you want. Nice, it will go to is Pexels. You can download images free for use. So pexels.com or you may have your own images as well. But yeah, go ahead to pexels.com if you, if you choose. Now, what I'm gonna do is I'm gonna go ahead and I'm going to click upload an image. And let's go ahead and go for a start with mountains. Now when you click to open these images, few things to consider. If you can't use a photo of a person without their consent. That is important as a rule of Darley. And also it's gonna get you to clock. Now can skip cropping. And that's a little bit of a better thing at the moment. But for this purpose, I'm just gonna go ahead and I'm just going to crop to hear. Because you can't generate variations if you have cropped, which is what we want to do just in this quick lesson. We're going to do that. I've uploaded this and I'm going to say, I want variations of this mountain. I don't quite like the layout of the peaks, but they're so go ahead and generate some variations. And again, what the software is gonna do is it's just going to take that image, tried to recreate it in that style. But like I said, it's going to try and generate variations of it, but in the same style. Okay, So this is what we've got. Here's our original image, and here we've got a little bit of a different layout. Different layout again, different layered. And you can see it's also varied, the trees as well and the fog. So you've got some quite different variations there. So let's try that now with an image which isn't, with an image that isn't an environment. I'm gonna go ahead and do it to this football. Now, let's just click Crop. And let's see, generate variations and let's see what it comes up with this time. Okay, So this is quite interesting. It's definitely doing the same sort of thing, but generating some real differences. I mean, I do, actually, I prefer this image to this one. Although I do like the little bit of light in there on the slide, the ball, but I much prefer the background here. This one's not quite working, if I'm honest, the texture on bull. But to be honest, this is really, really nice. I'm liking this, maybe just the white, but there isn't great. But this is still quite powerful. So let's go ahead and let's up the ante a bit. Let's choose another image. And this time I'm going to go for an image of, Let's go for a ship because the ship has got water in it and I want to see how well it deals with that. So let's go ahead and just crop and then click works. We want to see how it deals with the reflections. So let's go ahead and generate variations. Ok, so this is where it starts getting really cool. These are the variations is done and it's not done too badly at all with the, with the reflections. So let's have a look. That's not a bad reflection towards maybe a bit over the top here. This isn't bad at all. Maybe a touch. Not great for them, but on this side it's almost perfect. Here. Again, not too bad. And again by here, not too bad at all, a little bit over the top there. But this is really, really solid what it's given us here. So it can clearly handle, it can definitely handle reflections. Some areas, maybe it's not too great, but even if I look at this original one, then this is actually the photo. It looks a bit weird there, but that is actually in the photo. So to be honest. Yeah, this is all kind of really very, very good. So let's go ahead and let's up the ante even more. Let's go ahead and do one with an animal. So I'm gonna go ahead and do a dog and cat to animals at once. Let's click Open. Let's go for this. Let's click Crop and generate variations, and let's see how it handles. Okay, so I think we've got some interesting ones here. This is not too bad. It's not done well with the faces like we know. This isn't too bad again, but it's just the face isn't. That seems to be being a bit weird. This one isn't too bad, but again, just the neck and the face isn't great. So obviously we found a little bit of a weakness now. But again, it's not too bad. You can see what it's doing in the shadows or work. And if you look at the shadows for each of these, the shadow is working great just as it is in this original image. So you've got to, you've got to give it that. I've got one more image. So we may as well go ahead. So let's go ahead and upload the image of the squirrel and see how it deals with that. I'm really interested to see this because we've got a lot of good detail on there. So let's click Crop and generate variations, and let's see where it goes. Okay, So it's not done too bad because we've given it a clear idea of what their face looks like. I think that's what's helped you as the cat and the dog. The dog had a dark face and dark eyes and you couldn't, there wasn't much distinction, but this has done quite well. Note the shape of the tail being hugged to the back of the squirrel. It's really kept that shape going throughout, almost rotated it but by here and the hands are also in the same place as well. So very, very similar photo but you can tell that you can tell the hands aren't quite there. They're not quite perfect as they are in this one. I don't think you're going to manage to keep that. Here. They're not, they're just not perfect. But again, it's not done at two terrible job. These are certainly variations of this first image. So that is how you can generate variations of existing images. Now in the next lesson, what we're gonna do is we're gonna look at replacing part of an image. And we're going to start off with the ship photo and we're going to replace that with a different boat or ship. And we're gonna do a few examples of these. So awesome, I look forward to seeing you in the next lesson. 9. How to replace part of an image in DALL-E 2: Okay, Welcome to this lesson where we will be replacing parts of an image again using the Edit function. In the last lesson, we just created variations of images. But today we're going to start using the eraser and just replace in certain parts. So I'm gonna go ahead and I'm going to click Upload Image. And I'm gonna go ahead and find, let's start off with the ship and Angular click Open. And let's crop it just to this little area by here. Let's click Edit Image. I'm gonna get my brush and I'm gonna make it a bit smaller because I don't want to erase too much, but I'm going to just erase this bit here. And there we'd go. And I'm also going to erase the reflection as well just because I don't want it to include that. And what I'm gonna do is I'm going to type a yacht on the water. And I'm going to click Generate. And hopefully it's going to go ahead and it's going to start generating something to fill that gap. Maybe I should have gone with a smaller boat. I should've said a sailboat, but we'll see what it generates in a moment. Okay, so this is what is generated. And I'll be honest guys, it's not too bad. Very similar to earlier when we did, um, variations, but if we have a look, we've actually decided whether we want a different boat there now. So it's just focusing on changing the boat and keeping the background. So we could go ahead and we could click Generate again, or we could save whichever image out we need. It seems to have given us this weird little line downside, and that seems to be some sort of anomaly. So what you do is come back to the original and maybe do variations or click Edit again. Maybe extend this little raised apart slightly and say, let's say a boat and see what, let's see what a boat gets us and click Generate. Okay, so we've got some slightly different stuff. This not looking too bad at all. Let's have a look at the next one. Not a fan of that one. I'm not a fan. But this run again, it looks a little bit watercolor, but it's not too bad. So let's go and do a little bit of something different there. I'm gonna go ahead and I'm going to import an image of this rose. Now we've got, let's just go ahead and crop. And let's click Edit Image. Now we've got a vote here, but let's say e.g. I. Wanted this to be a sunflower. Well, let's go ahead and just start painting the area we want to get rid of. And I'm happy to keep a little bit of the stalk, but you can bring the brush down if you wanted to erase some of the stalk. Let's go and type sunflower. And that's the only word I'm gonna give it because it only needs to think about this area. And let's click Generate. Okay, so here are the examples has given us. They're a little bit quirky. Actually. Let's go and have a look through. I think this one's probably the better one. What I would probably end up doing is either generate and variations of this one, or perhaps I would do it again. So a single sunflower. Let's click back here. So let's click Edit again. This maybe give it a bit more room. And let's go a single sunflower and click Generate. Okay, you guys, so we've gotten us looking a bit better now. It may have been that I gave it a bit more room to work with or that I typed a single sunflower, but we've got something that looks pretty good. And then you've got this next one as well. Don't mind the lighting in this thrombocytes and lockers even rebuilt that shadow in the background as well. Is doing a good job at rebuilding that. Because if we look here, we've raised some of that shadow. So again, we've just, I don't like this one, but what we've done is we've just completely changed what the image was from. I think it was a rose, isn't it? And now it's a sunflower. So let's go ahead and do another one. Let's upload an image to edit. And I'm going to go ahead and I'm going to pick perhaps, let's go for coffee table and I'm going to click Open. Now we've got a bit more complex here. So let's just crop it to here, click Crop. And let's go Edit image. And let's say I wanted to get rid of, I loved this sort of setup, but I want to get rid of the laptop. Well, let's go ahead and let's just very quickly come through and erase that act. And then the prompt which we're going to use is probably something going to be coffee table with tea and cakes. So let's go ahead and just erase this. Now notice that we're raising a part of the flowers behind as well. So it's also going to have to consider that it's got quite a lot of work to do here. So let's go ahead. A coffee table with cakes and tea. And now let's go a marble. Let's just give it that or let's not, let's not go for a coffee table with cakes and t. Let's click Generate and see what it does. Okay, you guys, this is awesome. Let's have a look through this. So remember we had a laptop here to begin with, and I didn't tell it it was a marble in the prompt. I didn't end up saying that it was a marble copy table, but it got it. It looked at the texture and then rebuilt it. So let's have a look at this one. That one's not too bad against rebuild in the background as well. It knows that it's out of focus shrubbery. So that's really good. That one's not bad, but I don't know if though the cakes, to be honest. And yeah, we've got some pretty cool stuff here. So let's say again, I like this one. Let's maybe do some variations. Maybe we can take this a little step further and see how much better it can do. So let's have a look and see what that does. Now actually, because it's doing a variation of the whole frame. Interestingly, now, it's actually generating variations to the whole entire thing. We've got some very, very different over here. And again, very different environments about maybe something that you want to do. But again, what we did if we come back in our history is we had a laptop and now we've completely replaced it. So that's just a very quick example of how you can just replace parts of an image purely using daily. I'm sure if we typed something like an astronaut on a spaceship with that a raised, it probably wouldn't end up looking great because it's just gonna be too wild and two out there, although it would be fun to give that a try. Let's give that a try. So I'm gonna go and click Edit, and I'm sure this isn't going to look good, but it just be quiet fun to do it. So edit and astronauts on a space ship. This is just, just for laughs to be honest. Let's see what wild stuff it creates with that clumped, You know what guys? I think that's actually pretty good. I did say on a spaceship and it hasn't got that. I think it's struggled. But I mean, that's not bad. That's really not bad at all. Let's have a look at this one. Quite funny. It knows that it's got to rebuild the background. See, there he's kinda fallen fluid does actually not anywhere near as bad as I thought. But if we look back here, just, just, the thing we're thinking about is where we're placing an element. We replaced the ship with different boats. We replaced this rose with some sunflowers here. We replace the laptop with this stuff. So this is how powerful dahlias, you guys. So in the next lesson, what we're gonna be doing is we are going to be taking an image and I work actually going to be removing the environment. So I look forward to seeing you in the next lesson where we'll further build on this. 10. Changing the environment of an image in DALL-E 2: Okay guys, Welcome to this next lesson where we're going to be taking an image of something removing environment. I'm placing in a new environment, all with Dali. So I've got an image of a black car, and I'm going to hook that up here. And what I'm gonna do is just click crop and then edit image. Now you may want to pause or I'm actually going to fast forward. While I do this. I'm going to speed this up because what you wanna do is you want to try and be as close as possible. You want to paint out the background of whatever it is. Now this could be a person. We will be doing an example where I used a photo of myself next, but we're going to start with an object. So what you wanna do is you can zoom in as well, and you can alter your brush size as you want to get as close as possible. And you want to really just paint out the background of whatever your object or person is in. So I'm going to speed up mine as I do that. Okay, there we go. I painted out my car. Now what I'm going to type in is a prompt, a black car on a race track. So let's say that we wanted this on a race track, a car, black car on a race track. So we've given it a black car. Let's go ahead and click Generate. Now perhaps what it's looking for. Perhaps it just needs to say a race track and maybe it would do the rest, but we'll have a look and see what it does. Okay, so this is not bad at all, but I just click through here. Not bad. It's given us a race track. This is interesting. Maybe let's give it a Formula One race track. On a Formula One race track. I could give it maybe Silver Stone at an adolescent, let's just say a Formula One race track because it's kind of getting there, but it's just not quite getting it. Right. So let's just go ahead and click Generate and see where does this time. Okay, So we got some other things. This term, Let's have a look. I mean, it's dealing, this is a little bit weird up here, but it's dealing with the shadows. All right, not too bad. Let's go and have a look at this one, again, dealing with the shadows. Not so well, but let's have a look. What we got over here. And over here That's really interested in, is very, very sci-fi. I'm going to go ahead and give it one more go. I'm just going to say a car on a race track. I'm just going to maybe try and simplify it a bit and click Generate. Okay, so I did actually give it an additional thing here. I said a photograph of a car on a race track. Star should come back with stuff which is kind of okay. Now what you could do is you could go ahead and think alike certain parts of this, but let's go ahead and add it. And let's start drawing out the stand and something like this. And maybe you could almost say right, okay, brilliant. Let's use the same prompt and let's generate again, but I want to keep this part over here. So let's see what that does. Okay, so you can see from doing that, but we've managed to keep the foreground like earlier. We've actually managed to keep, sort of remove the background. But what we can do is let's kinda like I said, edit this, draw around again. And we could put a prompt like a vase in track stand, a photograph of a race and stand just to kinda see what that gives it. Okay, and we've got something look in a little bit different. I think I actually do prefer the sun, but you've got the standard, but they're, perhaps what I could have done is just drawn the standing in a certain area. It's tried to do the stand all the way across, but you can see what we're getting at. I think you can, with some images, you can kind of get this to work with some, it may be a bit harder. Let's go ahead and upload a different image. And this is gonna be a photograph of me now because you shouldn't upload photographs of people who don't want to be uploaded. It's kinda get what I mean, You got to get permission. So I'm just going to crop myself, but here, just clip clop. What I'm now gonna do is I'm gonna see if I can place this image of myself in. I don't know, It's more sunny environment, so let's go ahead and click Edit image. And you can get any photo of any person's not the highest res because it's been downloaded from Facebook. From my Facebook, obviously. But I'm just gonna go ahead and try and cut myself out and I'll speed this up for you and I'll see you in a moment. Okay, so I've done a really terrible cut out of myself there. It is quite difficult to use the brush tool. But what I'm gonna do is I'm going to put myself in an environment. I would say a photograph of a man, the heart. Okay, So I'm just gonna say it's very inside lighting. I wonder how it's going to deal with that because it won't actually affect the lighting on my face. I don't think so. Let's go ahead and photograph of a man at the park and click Generate and let's see what it does. Okay, so this is really interesting. This is not a park. I've no idea what it's trying to play out here. That looks like hotel room. And yeah, let's let's go ahead and change this up a bit. Let's just put a man at the beach and kept generate. So this is even more interesting. I have no idea what I'm supposed to be held in here. That's not a beach. I think what may be happening. I think the lighting on my face maybe maybe throw in a bit. So let's go ahead and put a man in an office, skyscrapers to see if it gives something better. Let's click Generate. Okay, so what I've done is I've gone with a man in an office and I've tried a few different things and we could probably do some variations of this. But again, environment stuff. When it generates a whole new environment, it's not too bad. I mean, you've got to consider, again, this is a neural network and look at the shadows. It is starting to kind of understand that. Let's have a look at another one. Obviously it doesn't alter the lighting on my face. Let's just click another generates to have another quick lock. But the shadows, it is kind of it is understanding that my face is lit up. The shadows should be happening behind me. We've got some other ones, but here, again, I don't know what that is on screen that looks like a gun. But again, we have started to replace the environment and I think if we do stuff a little bit more full-on, so let's say a man in a theme park, it does start to get a little bit more wacky. So sometimes what you asked for ends up being, you don't always get what you asked for with AI. That's obviously going to be quite a common theme. But yeah, it takes a lot of trial and error. I mean, have a look at how many times I've done attempts in history borrower here. And again that can eat API credit is quite fast. But interestingly, this isn't bad. Neither live tin doesn't match perfectly, but it's not terrible. Again, not too terrible. This one isn't bad at all. I think it does take a lot of trial and error. But again, you can remember back to the car one which we did earlier, that also took a lot of trial and error. So we did this and then loads of different environments and I think we ended up with something like this up here, which isn't too bad. So what I'm gonna do is, and the end that lesson there. In the next lesson, we're going to look at how we actually add an extra person into the photo. So I look forward to seeing you in the next lesson. Cheers. 11. How to add a person to a photo in DALL-E 2: Okay, so in this lesson we're going to look at completely adding a whole new person to an image. Now, this is quite experimental. I don't know, a few times before and it can take awhile. Which means it can also take a few, quite a few credits of your Dolly. But it is, once you do get that it is good. I may jump into Photoshop for this. This doesn't mean that you have to just, I may jump in there just to do one or two quick things. If you really wanted to get something that works really, really well, it would require some sort of additional photo editor. I don t think Dali is quite a bad pointer, could do it all yet. That's not to say that it's a requirement. I'm just showing you what Darley can slash can't do. But yeah, I'm gonna go ahead and upload an image. And this is another one of me when I was on my way to a wedding. So I'm just going to go click Crop. And it didn't need a cop anyway because it was swear, but I'm just going to click Edit Image. And what I'm gonna do is I'm going to come in and I'm just going to start painting out where the person who I'm going to add would probably end up standing. So what I'm gonna do is I'm going to come in here and I'm going to go quite wide. And just like that. And I'm going to then give it a prompt, membranous get smaller brush. I'm just going to so that could be someone's head of air. I'm gonna go ahead and say something like man with girlfriends now, honestly, some of the results that gives a quite hilarious you have to, you have to bear with it. So I'm gonna go and just type man with girlfriend will probably end up happening in this little video is I'll probably end up skipping and fast-forward in bits because I'm probably going to try many times and quite a few different key words. I've got some brilliant in the History tab right here. I've got some funny ones that I'm going to show you from earlier. It was wild. Let's have a look. Okay, not too bad. Somebody is playing light and seek with your bare in mind. You've got quite a complex background as well. You've got these paving stones which need to match. You've got the foil edge, which is complex. You've got these struts in the fence which would need to match as well. So just bear in mind that that's kinda going on as well. So again, faces just end up being hilarious. That's not too bad. Again. Yeah. Okay. Just just e.g. earlier, I was kind of this this is another image. It was let me let me tell you honestly, it was it was wild. Um, anyway, we're going to come back to these ones. I'm probably going to go ahead and click Generate again. What I may do when I find one that I like is a may try and generate some variations, are just arrays the girl's face. So I'll maybe just give me another face. Okay. So I'm gonna kinda go through quite a few times and it may fast-forward until I get a basis which I'm happy to build upon. Now, I've done many variations and I think there was a point to be made that Darley isn't the best at this. Let's have a look at some of the results. We've got this and then a quick look. Again, faces aren't great. Now we have got that additional software which may help, and that will probably end up pushing that software to the limit. This one's not too bad, you know, kind of integrated well and may have a look of building on that one. Let's have a look in here. It almost looks like some sort of expressionism art or something, doesn't it? Let's have a look at these, potentially that one, but again, not a kind of integrated well puffy feet. And let's have a look at this. Yeah, let's have a look at this one. I'm gonna do is I'm going to try and edit this again. I'm going to go ahead and I'm just going to try and really draw in the face. And possibly I might try and do the feet as well. But let's have a look at generating some sort of Thetis time. I did also put a comma and then the word couple at the end to try and get some different variations. But you can probably see that, yes, we can add a person to a photo with AI. But if you have the ability to go into Photoshop and make further adjustments yourself, if you've got that skill set, then I think it's at the moment, it's probably a needed skill set to really pull this off. Let's have a look now. This is starting to get a little bit better. Okay, so let's come across starting to get there, starting to get there. Maybe just maybe that GFP GAN that we looked at earlier in the course will help. So let's say e.g. we're happy with. Let's go for that was the original. Let's go for this face perhaps. For actually I think that's quite small base. So I think that will go for that one. Let's see if we can just picks the legs. So let's just go and. Let's also come across there. Okay, let, let's see if we can increase the believability of what's going on down here. So I'll just click Generate again. Okay, so we're starting to build upon something good here. We've got this not overly bubbly, believable. But maybe e.g. this isn't too bad. Again, if you've got the ability to do stuff in Photoshop, that'd be cool. What I think is really key here folks is look at the shading on her face and her body that they know that there is, the AI can tell that there is a person here. That's what's really, really good. Perhaps not enough people there, but let's go back. And yet again, you can buy that. So again, this is really, really good. I think I'm just going to go maybe an in real-time generate a few more. But I think I'm gonna go for this. Okay, The next thing is let's go ahead and download this image and we're gonna go back to our GFP gamma earlier. Now, I could just plug this into GFP gamble. I don't want it to affect my face. So this is where if you do have something like Photoshop or something else, It's good, but it's not needed. It's not an absolute requirement. But I'm going to do something. I'm going to show you something here which would make this just that little bit better. Now, let's just open this up inside of Photoshop. What I'm going to do is I'm just going to crop around this girl's face. Just then, but here. And I'm going to click, Okay, and I'm going to go File, Save As, save a copy. And I'm just going to go to my desktop. Girl. Face, PNG, click, Save, and click. Okay. And what I'm gonna do now is I'm going to undo my crop, okay, so we're back to normal. What I can now do is I can go to GFP gam. Click here. Go to our demo. Let's go and click here, sign-in if you need to, like we did earlier. Let's go Gil face. Yeah, I think so created a minute ago. Click Open. And let's just go ahead and click Submit. And hopefully now it's going to do a really good job at re, facing the skill. Perhaps could have done a bit better, but it's not too bad. I mean, it's definitely better than what we had if you compare this side to this side. So what I'll do is you could have a look at different version perhaps, and click submit and see if you get different variations. So let's just let it run again. Actually, that actually turned out better than I is still a bit dark, but I'm going to, for now for this, I'm just going to roll with it. So I'm just going to click Download. And again, I didn't want it affect in anything else. So what I'm gonna do is I'm going to go here, bring up my Photoshop, bringing this in. And again, you could use Game Boy, any other photo editing platform. So angular now convert normal that. I don't know if I drag this across into this document and just start scaling this down to roundabout. I'm going to do this really roughly. Round about some In, there we go. I think I'm about scaling a bit off. I should've kept the scale into one within that software. But if I turn this on and off, you can see that the face, Let's just maybe turn the opacity down slightly so that we can scale this, right? And something like that. Let's just go for that for now. And you can see that the face is a lot, a lot better than it was before. Especially that are there. So there we go. Now that is obviously quite hard photo because we regenerate in a full person there. You may have noticed is as we got somewhere where we were sort of happy, what I started doing was I started a raise in it bit by bit. Now, we, first of all, we come back to the history. We started off with this. Then we went to loads of different faces. And arguably that wasn't too bad. Then we started building the legs, then getting different legs. We started correcting it bit by bit. And I think that's the process you're going to have to use. Again, if you do have Photoshop skills, you could come in and you could almost come back at the original image of me in here and sort of build back in my original shoulder. And again, it would just start to come together some more. I think Our Delhi is great. I think you do have photo editing skills. It is certainly a plus. So I'm going to leave it there for that lesson. In the next lesson, we're going to have a look at expanding the edges of the photo and taken a photograph or a piece of artwork and expanding it. So e.g. if I had this photo, this original, whoever the original is gone, the original, and then started expanding the photo to build a new environment. That was shot in my back garden. And if we wanted to expand that to give more, well, that's what we'll look at in the next lesson. So cheers for tuning into this lesson, and I'll see you in the next session. 12. Expanding the edges of an image in DALL-E 2: Okay folks, welcome to the final lesson of this course before I set you your final project task, which will be related to this. So what I'm gonna do is I'm going to show you now how to use something called Generation frame. So I'm going to click on upload an image, and I'm gonna go ahead and select portrait of a woman. Okay. I'm just gonna maybe skip cropping. This is experimental. It's more, it's more beta mode. I'm just going to click here. Now what I'm gonna do is I'm going to add a generation frame. And you can do that by clicking Add generation frame down here. So I'm just going to go ahead and do a cube. So even though I've got a portrait photo, but I'm going to go across but here and just go somewhere like that. And I'm gonna go countryside. And basically what this does when you click Add generation frame and then you give it a prompt. It looks at the artwork and then it looks at your prompt and it tries to generate countryside, but based upon this art style. So this is a really cool way of extending an image. Now interestingly, as popped up with a different art style here, but let's click a course and see, yeah, okay, what else we've got now this is more what I was expecting. And it's done that for the three of them. So weirdly, this first one is a massive anomaly, which honestly I was not expected at all. And I was a bit like what? But this is definitely what was expected. And I like that. It's really, really a thing. I'll prefer that. Or definitely that I'm going to click Accept. I'm going to go ahead and I'm going to extend this now you have got that seam down the middle. So I'm just going to drag it and hold spacebar to click and drag. And I'm going to add another one, generate. Now, there have been people who have made really interesting cool panoramas with this. And I'll show you them in a little bit. But let's see just how far we can push, push this with just this single prompt. Okay, so we've got some more stuff. I'm loving that. I love that as well. It's got the little house and that house. I'm loving it all. I'm going to go with this one and I'm going to click Accept. Now I don't want to leave out the other side. So let's go and do it on the other side. And I'm going to have the same just run down there. So hold spacebar to click and drag, and I'm gonna go ahead and click Generate. Okay, so let's see what options we've got here. Quite more scribbly because maybe, perhaps because this side of the original image was more pencil work in scribbling. Let's have a look. Yeah, I don't mind that. I think I'm going to go for this one. And you can see it's starting to adapt a different by zoom out holding the control wheel and scroll in that starting to get a different art style this side. Well, let's carry on this side. And let's click Generate. This is really cool. I love that. I almost want to click Accept now. But then again, that's cool. Now I think it's this one. But obviously we can do the sides as well. Let's do the sky. So let's just keep on going on up here and click Generate. Okay, Let's see what we got. Definitely not. I mean, that's quite interesting, quite abstract. As a mirror almost looks like maybe I'll go with that. But again, you can just kinda do this and do this to your heart's content and make a massive, it'd be cool to me, the massive portrait. Let's go further down here and build up maybe some of the bottom down here. Okay, so this is where it's starting to get a bit wacky. And perhaps if anything, I'd go with that one. Now interestingly, let's have a look. What else we've got our generation frame, upload an image. Let's try a razor and let's just try a raisin, just that part because I don't know why that bold thing is there. Let's just try and click Generate again and hopefully then it'll fill that gap. Obviously you can see what I'm doing now I'm building upon the prompt scales which we had first of all edited and added and removing things. And this new skill of generating a frame. But let's see what it fills this frame width and hopefully it'll be with something good. Okay, there we go. That's I don't know what the blue thing is, but that's not bad. That's pretty cool. Okay. Yeah, happy with that. Happy with that. Cool. So I've done this with a painting. Let's, let's try and do it. And this will be, this will be new territory for me. Let's try and do it now. Have done it with the Photoshop, but yeah, let's go back to Dolly. And leaving this page will discard the changes. Yeah, I don't mind. I'll click accept whatever. Let's go back now to dilate. Okay, cool. Let's just go reload. Finally, upload an image and let's go me in a suit that one familiar. Okay, let's just click Crop, edit image, and add generation frame. Okay, let's, let's just go and add some stuff by here. And what we're going to type, this type, countryside. Yeah, let's, let's try and do countryside runoff. And let's click Generate. Now, it's gonna be interesting to see whether it does keep that wall going. I don't think it will, but let's see. Okay, So that's interesting. I did keep the wall go and let's have a look at the variations are like that, that feels a lot more realistic as does that. Actually that one does even more. So I'm going to accept that one. And perhaps let's go ahead and just see what it does kind of up here. And click Generate and let's see what happens. Okay, so let's just go ahead and scrub through. I think it's the first one I liked, but here, let's do one more, just over here. And let's click Generate again. Okay, Let's see what we got. I'm not a fan of that. I don't mind that. I would perhaps generate this again. Okay, perhaps what we would do is click Cancel. There we go. And now I'm going to generate again. There we go. Let's see what happens this time. Okay, not too bad. It's a bit weird. I would probably and I'm being quite picky with this one down here. And I'm just gonna go for argument's sake, just to end the last name we give them. Definitely pick yellow. I'm going to go for that. Okay, cool. And click exception where you can see I've just basically starts to extend that image massively. And I could probably give it more trees up here. Let's just do that one more. Okay, Let's see what we got. I mean, one thing I haven't been doing is I haven't been changing my prompt. And arguably for over this side, I should have said something different countries, so it's quite vague. So I could have probably given it a lot more. I'll just kind of I'll leave it there for now. I'd like e.g. up here, countryside and trees and stuff is probably gonna be okay, isn't it? But anyway, what I'm getting that folks, is this is another feature extending an image and painting in more environment to your original image. So I hope you find that lesson really useful. Hope you found the course useful, but it's not over yet. There is a project task, a final project assignment in the next lesson. So tune into that and I'll see you there. 13. Final Class Project: Okay folks, you've nearly reached the end of the course, but as time for your final class project, now what I want from you in this class project is ultimately I want you to end up uploading another image. But using the skills you've lived in the last lesson where you take a source image or one that you generate and you expand using the generation for you when you expand the edges. Now, you'll be seeing on screen a few panoramas which people have managed to make with this. And it's really, really cool. So your class project is to take an existing photo and upload it or generate one initially and use the Add generation frame to generate a really cool panorama. Now you can use the same prompt throughout or you can use different prompts. That is absolutely up to you. But what I want is an image uploaded. We've used the ad generation frame to create a really cool white panel, violent. I really can't wait to see what you guys create, make sure to upload it. And most of all, thank you for joining me in this course. I hope to see you again soon.