ComfyUI Flux Kontext Mastery: Create, Edit & Transform AI Images Effortlessly

1. Introduction: Hi, everyone. I'm Karen, and I'm super excited to welcome you to this course, Luxe context using CoVUI where we dive into the world of image editing and image generation, and this software is totally free within your computer. In this course, I will show you how to turn simple prom into stunning visuals, edit images with precision and unlock a whole new level of creative freedom. Or with tools that are accessible to everyone. So what we are going to learn together, we are going to learn how to install comp ui from scratch, we are going to learn how to generate your first image using prompt in copuIimage editing made really easy using flux context. If you want to change the hair, you can do. If you want to change the color, you can do. If you want to add wings to cat, you can do. If you want to add a dragon in your backyard, you can do. If you want to edit or manipulate your own images, you can do all the advanced features with simple steps. You can even do branding to products, add products or remove products or add logo to an AI character, you can do. Branding of products is really easy. Why flux context and CF UI, because they are powerful, flexible, and best of all, they are free. Who is this course for? This course is for freelancers, digital artists, and obviously, if you don't have any knowledge about AI generation, you can take this course, and anyone who wants to love experiment with new creative tech. What do you need? You need a computer, eight GB of RAM or 16 GB of RAM and eight GB or 16 GB of graphic card. If you have these requirements fulfilled, you can join in. I can't wait to see what you create. Let's get started so in the first lecture. 2. Install ComfyUI for Windows the right way: Going to install Kofi UI with a press manner because there might be some change in the step. So for that reason, we are stalling or deleting all the files of Kofi UI and reinstalling everything from the scratch so that you might not get confused because as the technology shift, we also have to update ourselves. So first, let's understand how to install Cf UI in a perfect manner. So first, we have to prepare some file to install that. For we have to go to. So here are the two links. So first, we have to install Git. If you have not installed the Git version, you have to install it, download for Windows. You have to install it. You have to download the Git version here. So you have to install Git 64 version. I will provide this link in the resources so you can find it easily. Simply click this. After that, you have to install this. Simply double click it, install it. So you have to go to your CompuI Github page that is CompuIlas desktop. So Compi has launched its client, so you don't have to install this environment anymore. So you have to down Windows NVDA NNS and SIS or if you have MacOS, you can download this one. You have to click on Download button. Once you click on Download, download will start automatically. Let's wait for some time. It is about 108 MBO files. Wait for 3241 facon. And now click on Install. Let's open it. Let's open CompuI. So here is our CompUI. We have already installed this kit, and now we are going to install CompuI. So let's double click it. So now Comp UI Window is pop up and it's very colorful. Let's get started. Now we have to select NVDA because I'm using NVDA graphic card, then I'm using NVDA Cuda device. And if you are using CPU and if you don't have NVDA, you have to select CPU mode. Let's next install it in the C app menu. Let's install it somewhere else. Let's install it in New folder, that is new Comp UI. Let's install it here, select folder, and install it there. App roaming app data program UI electron source. So here's the system location, Cf UI going to use. You will C drive, press next, migrate from existing installation. If you have existing installation, we can copy your existing file. So let's not copy any of the files. Let's keep it next because we have already deleted the files, update automatically. Uses matrix. Yes, we need to see what uses matrix mirror setting. Yes, Python install mirror. So let's make it installed, and it will take so much time and your Internet connection. It's going to be installed many bigger files for you, so it will take some time. You also have Python version 3.10 or latest. You simply have to go to Python Download and download Python. This link I will provide in the resource Sext, so don't worry about that. Install. 3. Update ComfyUI for Windows Portable Version: So let's quickly go over how to update Comp UI if you are using Windows portable version, not the texto one. So first step is open your Comp UI folder and look for the Cofire Windows ctable directory. Inside that, find and open the Update folder. Now, you will see three bat spies. Update Comp UI, update, CofiUI and Python dependency. Update Cofi stable. To update Cofire to the latest version, just run these 351 after one. Double click each file, wait for it to finish it, and then move to the next one. That's it. This process will make sure Cf UI and its dependency are all up to date. Super easy and quick. Now, restart your CUI again. That's all. See you in the next lecture. Thank you for watching. 4. After installation generate your first image: So once your installation is complete, another windows will pop up, that is getting started with a template. All these are examples you can use as selected, Haz a flux, Dave snail, in print, can dep redux model, image, control net, upscaling, video area composition, three, audio. Let's start from the basic image generation. Let's start from the image generation to loading the graph flowing nodes are not found. I automatically show all information in front of you and here is the checkpoint that they want us to download. Let's download this, wait for some time. Once you install it, you simply have to cut it. That's odd. Now let's run a prompt. This is first time we are going to run it in our software. Let's click here is the image. Wow, we have beautifully created our image. This is the first time that we have created our image. Now let's head over to our workflow and interface how we can use this interface and start creating our workflow on our own. You will learn a lot from this. Thank you for watching. See you in the next lecture. 5. Update comfyUI within the software: How you can update Cf UI. Again, there is another Ato update CFI, how you can do that. Here is the window that will pop up once you launch your CFUI. You can install now and restart CUI, or you can go to manager, once you click on Manager, update all custom nodes. That's all you have to do. Let's install it. Once you Clic on install, it will take some time to install or a startup task has completed. You have to click on Restart. Once you click Restart, CI will automatically relaunch itself. If not, you have to click on CUI, double click it. Now your CUI has been updated, wait for some time, and it will load automatically. Let's click it, Install now and restart Comp UI. It's asking me to relaunch it. ComfUI has an update and yes, its icon is also changed. Let's wait for CompuI and it's installing itself. It may take some time. Your ComfUI has been updated. 6. Introduction to Flux1 Dev Kontext: Let's start with what flux one contact table is all about. This is a powerful open source image editing model from Black Forest Lab. It can understand both text and images at the same time, which means you can give it instruction in plain English and it will intelligently edit your image. It's great at keeping character and object consistently, even if you make lots of changes. If you want to edit an image multiple times, things like face, pause, and layout stays stable. It's really cool. Now it's easy to use. So there are currently three main versions Pro Max Dave. Pro is for commercial use, Max is experimental version. Dave is open source, perfect for research and people like us, how to download and required models. These are the information that they have shared, how to download, and where to download. So we have to download diffusion model, VAE model, text encoder, and these are the folders that we have to download in each of the Pis. Each Pis we have to download in each folders accordingly. 7. How prompt work in Flux Kontext: Now let's talk about how to give instruction or prom to flux one contact tap. About the basic modification, if there is any easy change, change the color to that. That's all. Keep the wife, change the daytime while maintaining the same style of the painting, and if you want to transfer the style, name that style, transfer to dash art style. That's all. Be specific, preservative, preserve features step by step. There is something I call golden rule. The more details you give, the better the magic. Context Dave loves his specifics. Download your files, put them in the right place and get creative with your prom. Experiment, have fun and let your image run. Don't worry, this is just an overe. In the upcoming lecture, I will share the workflow with you so that you can just drag and drop and use them directly in the CFUI and start creating images right away. All the download links and all the file location, I will be sharing you in doc file. Don't worry about. 8. Download files in the right folder model and VAE files: You have to copy all this folder into the fui into your CUI user and default workflow. You simply have to copy and paste all the folders that you have received in Zip file. So you have to extract them here. After that, you simply have to go to your workflow and click on this folder, workflows. That's all. And after that, you have to repress it. Once you rephrase it, all the folders with the workflow will appear here, how many workflow they have. So first, we are going to explore the image generation using context. How we can do that? Let's install it. Simply, you have to drag it or simply have to click it. Once you click it, once you click it, it will ask you to load all these missing nodes. You can install, install missing nodes, apply changes, place restart, or you can do by doing open manager, it will show you all the missing nodes. Right now, we have already installed the missing nodes so you don't have to install it again and you simply have to restart it. All right, team, here's how you can get everything set up for flux contacts. Super simple and stress free. Now after installing the node system, you will need a few key files to get it starting that diffusion model, VA file, and other important model files. Where do you get them? Option one, check the PDF in the download section of your course. It has all the links you need. Option two. When you open the workflow and CompUI, look for the nodes section. Here are the direct links you can click to instantly download each required file. Put files in the right folders. After downloading, make sure you place each file in a specific folder. Inside your CofUI directive. Once you download it, you have to go to your CUI, in the models, here is a model diffusion model, and you have to download it here in the diffusion model. You have to click it and install it. It is around 12 GBOFle. We have to wait for it, and for the smaller version, you can use this or this variant. It is depending on the VM you have because it is for the 12 GBO vim and these are for the eight and six GBO VM. After that, you have to download T five VL. That is T pipe VL XSL encoder, Q GGUF. You have to download this clip Encoder from here. Once you click it, the download will start automatically and you have to move this file into the specific folder that is models and clip. Once you go to CUI, like we did before, once you go to CPI in your core folder of Copy, you have to go to models and search search for clips, here you clip, click it and you have to install it here. That's here. For the smaller version, you have to install this variant. It is for eight GB of frame or six GB of MM or four GB of frame. It is depending on the VM because it is for the 12 GB, at least for the 12 GB of VRM. For the VA E loader, we have to install it in the VAE folder. I recommend you to download it again because all these are updated file according to the CFI UI, so you can install it in the model and VAE. Let's find out in the models and VAE, here is the folder and you have to install it. I have already downloaded it. It is around three MBO file. After that, load scale model. You have to install it into the Comfy model is scale upscale model. You have to download this file. It is 63 Ambio file and you have to install it in the upscale model. About the nodes installation, we have already did from manager custom node manager or missing nodes custom nodes. You have to search for custom node. If there is any, you can install it. So they are custom node that we have to install it. Let's wait for our main file that we have to download it. After that, we will run CPUI and if we face any problem, we will work accordingly. If you ever unsure, just refer back to the PDF in the download section, it's got screenshot and step by step instruction. Or if you are working inside CompUI, those handy links in the workflow notes will take you straight to the downloads. That's it. Download, organize, and you are ready to start creating flux contact tape. See you in the next lecture. Thank you for 9. How to Install missing nodes: So we have installed UI EZ us and RG three. As you can see, from the manager, you can go to install missing custom node or you can search for custom nodes like CofUI and search for Z. I have already installed it, as you can see, lean, you have to check for it and install it. After that, you have to search for another node that s RG three C PUI. As you can see, it has a 2012 stars or reviews and you can install it from here. 10. Using Kontext model Generate out first image: We have already installed it. After that, now this is our first run of contacts day plus upscale. I have added the upscaler here. You don't have to worry about that, you can ignore it and if you want to upscale it, you can upscale the images further. Or you can do that. You have to simply keep your seed same that all, you don't have to randomize this or increment it. We are working with same seed. You can change it. It's up to you, just click and change the number. That's all you have to change and your image will also be changed. Right now, let's change the seed to 5050. This is the seed I'm using right now and let's run prompt first and I will explain you the workflow how it's working right now. Let's run in and wait for some time. So this is the prompt and adventure as Dog astronautipping, and this is the result. Result is really good, as you can see, how this workflow works, first, it's load the diffusion model. As we know the diffusion model is our flux tape context this time, it's not flux, not flux SNL and any other checkpoint. There is no diffusion model. Right now we are using Dave contacts, FP eight. Now, for the dual clip encoder, this is the same as for flux that we have used before and now, it does not have any negative prompt, as you can see, it does not have any negative prompt and it has the positive prom, this is it for the negative prompt, they have used another condition that is zero out. This is the one that is new. They have shared in this workflow from Blackab team and after that we have a flux guidance. Before we have used flux guidance of 3.5, but right now it's working on 2.5 and it is recommended to use the guidance of 2.5 for contexts. For K sampler, this is the brain which encode everything and the step that we are using for context is about 20 for the 20, the result we get for flux context is really good. I recommend you to use a step 20 in VM used, it is basically used for VRAM to clear out the memory that we have used before so that our workflow works seamlessly and perfectly. You can enable or disable it the upscale, yes or no. Once you click it, it will automatically generate same result in another run, you can simply enable it and it will run automatically and upscale your image by four X. As you can see, the load upscale model that we have used is four X, so it will upscale by four X. We have used 1024, so it will multiply by four. So the result we get is about 4,000 pixel of image, which is a very high quality. Right now it is at d quality of image, which is square one, as you can see, 102, four, 1024. Let's disable it right now and jump to our next workflow, which is image generation using Context GG. We are going to use same prom here like we have used here. Let's jump to this 11. GGUF Kontext missing nodes installation: There, welcome back to this course again. Right now, let's copy this prompt and click and open this Wblow. Right now, we have no missing load, but we have Unit Loader, GGF, and Dual clip Loader Gf. These are the missing file. We have to install it. How we can install it, as you can see, flux contacts Dave at GGUf. We have to download this file from here. Once you click it, it will download automatically and the file is about 12.8 GB. After that, you have to install TX encoder GGUF. That is T five VF quight. You can use smaller verson if you have a lower VM. Right now I have 16 GB of Vim, so I'm using our qu eight. Let's click it and start downloading it again. So we have a 12 GB of GGUF and encoder is about five GB of encoder. We have a encoder of five GB and context is 12 GB. After downloading, let's jump to future and after you have downloaded this, you have to run it because missing node type node. We have some error here. Let's find Unit Loader GGF and missing node. We have some missing node here. Let's go to edit repress node definition. Let's do it and let's run it again. Okay, the issue is still not resolved. Let's unit Loda GGF, dual clip loader GGF. You have to find install missing custom nodes. Yes. We have two custom node that is not installed here. We have to install it one by one. Latest select install this one also. Install it. Now, restart again. 12. Understand the flux Kontext workflow each node: We are going to understand our workflow. That is flux contest tape plus upscaler. In this workflow, in CFUI, we will walk through every node you see on the screen, how each piece fit into the magic puzzle of AI image generation. By the end, you will not only understand what each nodes does, but you will also have the confidence to experiment and create your own workflow. It's going to a very important lecture. I break it down in a very simple step so you can understand it easily. First of all, you can see load diffusion model. From the left side on upper left and side, you can see load diffusion model. What it does in a simple language, imagine this as choosing a brain for your AIRtist. What it does, this node loads the main model. That is flux one Dave context in this scenario, or you can use flux one Dave or flux one SNL. If there is any checkpoint, you can use that. In the future, if there is diffusion model created by black labs, you can use that too. It will generate your images. Why it matters. The model is like the style and skill set for your AI painter. Choose wisely for different artist result. Here is a small tip. You can swap models here to get different vibes from the photo realistic to cartoonist, depending on the model that you are using. Now let's go to different part that is dual clip loader. Think of this as giving your AI two sets of glasses. To better understand for your words. The prompt that we add into prompter or in the clip encoder, it will understand that English. That's all or any language if it can understand. What it does, load two clips model that is clip one Sapner and FSCLF 16 Sapner. As you can see on your screen, clip model help the AI understand your text prompts, why we are using two clip encoder. Sometimes using more than one clip helps the eye capture more nuisance or style from your prompt. One is understand the English that the prompt has specific in it, or if there is any style, the other will work on it. That's what it will do. Playful analogy. I like having both a poet and a scientific interpretation, your instruction for the AI. Now, let's understand clip text encoder, where we add prompt or give prompt to AI. It basically converts your written prompt into a language the AI can see and feel and understand. In 010101. Now let's move to other part that is flux guidance. What it does. First of all, this is your AI personal tener. Keep it on track with your vison. It guides the image generation process using a prompt and a clip model. The guidance value control how strong the AI stick to a crank it up for more later results or just down for more creative freedom. That's all. There is one fact about it. Too much guidance can make the AI stubborn, sometimes a little freedom leads to happy accident. What it does mean, it's only tell us that if you increase the guidance, the AI will stick to our prompt and give you a result that you might never think of. Also, if you give a loose point that is 2.5 to 3.5 flux guidance is perfect for image generation so that AI can think on its own and give you the desired result. So now let's move to another part that is conditioning zero out. A result a reset button for creativity. What it does, clears or reset any leftover condition for previous run, ensure your next generation start with a clean slate. After that, here comes load VA, the AI translator for turning dreams into pixels. This is where the magic happen. What it does, loads the variational encoder, that is VAE in short, which helps decode the AI internal image in something you can actually see model used aa dot Safner. This is a model that we have used, VAE safe. If you want Suba change the look of your final image, so you can experiment to find your favorite. That is VA encoder, you can find on Civit AI on any other platform, you can find it and use it so that you can give us special test to your final image. After that, ampity as T three latent image. What it does setting the stage for your masterpiece. In a simple language, as you can see, if we use tant tenet pixel, it gives the ampute stage or amp space for our encoder so that our image will come in same exact resolution or pixels like we have given. If not, it will always give you in square size. It prepares a blank canvas in the AI internal language, that is we call latent Ispace. Width and height, you can define as your own. If you want to use it in real, you can create a real format or in the real resolution. You can do so using latent image. Batch size, how many images you want to generate at once. This is what it does. Now, let's go to K sampler, the conductor for your AI orchestra. What it does controls the actual image generation process. How many step randomness seed, and which sampling method to use. Let's understand case sampler because it is an important part that we have to understand. Imagine you are going on a road trip in a car, a beautiful AI generated image. Keampler is like the driver of the car who controls how the journey happens from start to finish. Seed is just like GPS coordinates. Just like entering a location pin into your GPS or Google map, the same seed will always take you to the same image. Changing the seed, you will end up at a different destination. That is image. After that, is steps number of turns on the road. So steps is just like number of turns on the road. More step are smoother and more detailed journey, fewer is step, a quicker trip, but maybe the image is less detailed. You have to simply understand once you increase the step, your machine has more time to create this image, fewer step, fewer details. Let's talk about CFG, classifier free guidance. This is what we have to tell the AI how strictly you have to follow the map. Low CFG, you drive freely taking side roads, anything you want, creative results. High CFG, you stick exactly to the GPS route that is prompt for accuracy. It basically helps balance creativity versus control. Now sampler, that is dular. This is the driving style. Different samplers are like different driving styles. Some are fast and aggressive like Euler. So are slow but scenic. Like DPM plus plus. You have to understand and experiment with different cellular or different sampler and all the others are balanced. You have to experiment with them. Each gives a different feel to the final image. Design, denoise, how much you clean the wind seal. Denoise what it does. The AI start with noisy blurry image like looking through a dirty wind shield. Once you clean the wind siald, the cleaner the image. Low denoise is equal so low denoise, more original randomness. High denoise, you clean the winds in mode, stick closer to the prompt and less chaos. Once you increase the denoise, you will get a random image. The AI will change many things in your image and give it a creative look. Once you increase the denoise, it will stick to the prompt and it will not change major things in the image. In short, K sampler is your skill driver carefully navigating the AI car with tools like Step, CFG and sampler to make sure you want to arrive at the exact image you are imaging. Here comes VAE decode that bring your AI dams into the real world. It decodes the AI internal image into the viewable format like JPEC or PNG. This is where the magic becomes visible or the prompt that you have given is visible. Clean VRM used, tidy up after the creative mess. It does simple thing that clear your video memory, that is your graphic card, I clear the RAM in it, that is used during the process, and keeping your computer running smoothly. It prevents crash slowdown, especially on long creative session. Save image original. This is the gallery for your masterpiece. This is where your image will save and the location that you have given to the computer or your machine, it will save automatically in the comp output folder. That's what it does. Simple. After that, enable upscaler. It's basically upscale your image in a simple language, increase the resolution of your image, increase the softness of your image. In the upscaler, once you go to upscaler, it does same work as the main workflow that I have explained to you. It just use upscaler diffusion model and dual clip encoder into the a sampler. That's all and the process is same. Once you change or experiment with other value, your upscale image will also get affected accordingly. That's you have to understand and it takes some time to increase the size of the image. One protip. Always remember the number of the seed, once you forget the seed value you will never get the same result to upscale it, you have to fix the seed number so that on the next run, the AI will generate the same image and upscale it because once you randomize the seed number, the AI will never upscale the same image once you run the prompt. It will generate a image again and upscale that new image, not the previous one. That's it for this lecture. I hope you understand the workflow in a simple language, and if you like this course, please share your review. Thank you for watching, see in the upcoming lecture. 13. How to upscale any image and Understand GGUF generation: Up installing it, you can find that it's now not showing any kind of red error. Let's run it Adventure Helgo Asnaut. Let's run this prompt, like we have used before. It is the same prompt that we have used before. Let's wait for some time. For the seed, we have used 050 right now. You can change it to any number. It's up to you. Here's our result, as you can see, it looks nice and really perfect. It is very clean, love it. Results are really Fine. And if you want to upscale it, you can simply enable upscale and simply have to run it. That's all you have to do, and it will take some time to upscale. Like run it. That's all you have to do. It will take some time to upscale it and if you love the result, you can move forward and if you don't love the result, you can try again with a different value. If you want a different type of result, you simply have to change the seed here. That's all. You can upscale your image. Also, you can generate unlimited images using Context, Cf UI, increase the size of it and I prefer you if you have a lower VM or lower graphic card or lower GB of graphic card, you can prefer lower version of GGU. Right now I'm just canceling it out. Execution has been interrupted and it will take some time to cut it out. Now, let's go to another one that is image editing and right now we are going to combine two images 14. Understand combine 2 images and What not to do: Some people, it might take different time according to the GGUfversion F you have used and about the main version of context. For me, the GGF version tos very long time, about ten to 15 minutes to generate image. Let's check it out. As you can see, it took about five 9 seconds to generate this image. So for this version of image, it took about another 1 minute it just generate the images. For me, GGUFs not working that much fine, but I think I have to use a lower VRAM version that is q six or our quarter. You can download it from here. You just have to click small variant and it will redirect you to GGUF or Gap versen of lower VDAM. You have to download it from here, you can apply anything. But according to the nodes in the model card, you can see that at least q five and larger version is better in giving the results in image generation. It's depending on the variant that you have. At this you must have a four GB of rem all lower, it depends on you. You can check it out. And depending and test it out. Which one is giving you a better result or taking less amount of time, it's good for you. You simply have to test or experiment it because I have not experimented on this workflow that much. I love the result of eight. It's fine for me. Let's move to another part of image generation. Let's go to workflow and library and go to image editing and go to combination of two image. I'm using this image that I have generated before and right now I'm using this image. So let's see how it will work. After loading the first image and second image, let's keep the prompt in a desired way. That is, right now, I place the text, a prompt that you can see, place the man and forks together in one scene, wipe being exact facial features, eyes colors and facial expression. The man is gently holding the forks in her arm like a father holding his baby. The scene style as elegant photography captured with a professional DSLR camera. Soft morning light with natural cidos Run this prompt. These are two images are being stitch now the case amp is loading and wait for some time, it will actually amuse you every time you run this prompt. Now our CD 30 pi, let's wait for some time. As you can see, this is the merge, and this is the result. I think my prompt is not that good because of the prompt, I give me the results like this, no problem, but at least I have given you the example. How we can change or manipulate our images. This is me. Please don't mind my outfit. Please mind my outfit here, and I love it. The results are amusing. I told you, every time you run the prompt on it, it just amuse you every time. How much time will it take to create for me? It took about 187 0 seconds. It took about 31 point 1 minute for me to generate this one image. Why? Because of the size of the image that I have used and after that, it just computing all the images here. Depending on the size, it will take some time for you, but the results are not. 15. Understand the prompt Combine 2 images: So right now we are trying to combine two images using CFUI with using the context node. I will walk you through what we went wrong with our Isa setup, show you the result, and then we will tweak the prom to see if we can get a better outcome. Let's jump in. All right. So here's our CUI workflow. We have got two images loaded in, one of an astronaut in the field of flower and another image that we wanted to blend in. For the context note, I use this prom. Fox is walking with the astronaut, keeping the model lightning and colored unchanged. Sounds simple enough. The idea was to have fox appeared in the scene with the astronaut, but keep the original lightning and color. Now let's look at the result. As you can see, the output is basically just a original astronaut image. No fox insight, so what we went wrong. Here's what happening with our prom. The prom tests the model to add a fox, but not specific enough about where or how the fog should appear. The model might just ignore the new element if it's not clear. Context not limitation. Context is great for conditioning, but it can struggle when you want to merge two very different subjects unless the prompt is super clear. Image conditioning. If you don't give enough visual guidance or use a the model tends to stick closely to the original image and ignore suble prompt changes. That is why we don't get the desired result. The fox wasn't eddy because the prom didn't force the model to blend the two images or make space for the new subject and keep the model lightning and color unchanged. Now let's fix the prompt. What we do now, let's change up the prom to make it more direct. Instead of just saying Fox is walking with the astronaut, now let's run the workflow again with our new prompt and see if we finally get that fox, walking with the astronaut. Stick around for the result in the next video. 16. Understand the Kontext size concept: I think if you have noticed that the dimension of this image is different and this image is different. But still, it's find a way to adjust both the images and give it a perfect results. How is it possible? It is because of this node, flux context, image is scale. Let's understand this node. Let me share what it does in the new workflow. As you can see, double click and add a load image node. After that, add image context, as you can see, load. Flux, context, image is scale. D click flux, context, image is scale, and after that, add a preview image. The word click Preview image, turn. Now, once you run the workflow, as you can see how it adjusts the size. Basically, what it does, if you choose a image that has a size of 1024 and run the prompt, the size you will get is 1024. But what happened when you choose different image, it will automatically change the size to its optimum size. That will work with context. That's all it does. In a simple language, it changes the size according to the need of our workflow and change it to 1024 to 1024 or depending on the latent space that we have created of different height and width and give us a desired result. 17. Perfect prompt to combine 2 images: You have to understand the prompt. Like the prom we have used before, there is no exact meaning of the prom. Right now, there's a meaning of the prompt. We have to understand the prom and create it like this. This is the one, this is the two and place both Astronaut and Fox in one scene where they are walking together. Wow. Amazing. The results are mind blowing. I love this. Now if you want to upscale it, you can do so. You simply have to enable it and run the prompt. That's all you have to do. Combine the two image has been done. Please experiment with it and share the result with you. Thank you for watching, S in the next lecture. I recommend you to save the workflow before you close it. Once you close it, a default workflow will load again and after that, let's go to amazing thing. That is image editing. 18. Generate image in any dimension: So now let's understand this workflow. How this workflow work. In this image, as you can see, it has a fixed custom size that the flux contact image is scale will give the output and according to the output, it will generate the image. But right now when you image rating custom node, when you use this workflow, it has a custom nodes. That is a amputee latent image. This is the extra latent space that we have generated here that will go into the latent image. But before that, the VAE encoder or the latent image was connected to this latent image. Right now we have changed the width and the height of the image and you can add it according to it. How we can do that, let's use another image and change the dimension of it. All these images that I have generated here is from Adobe Firefly. I just take the references from the Five fly. I don't have to use my brain, how many types of images I can generate. I take the reference from the firefly. You can do so. You can use Pinter firefly, anything you have, you can use or you can use Chat JPT perplexity. Is depend on you. For the ideas, I use Firefly in these lectures. Right now, let's drag this image here and let's change the width and height to 1080, 1080. 10 801-080-1080 80. This is the image and this is the pvment after the image and Latin image, we will get the result according to our squared image. Right now, let's add dragon very angry. Add dragon to the scene, very angry with fire in the Mouth. Let's run it and wait for the result. Image is not that bad, but I think image is very bad, to be honest. A red dragon to the scene very angry in the mouth flying above the sky realistic scene. Let's run it. I have changed the prompt a little bit. Let's see what type of results we get because this is really bad. So this is the result that we get right now and it is better than before because it's now flying to the sky. But yes, I don't like that much. 19. Add object in the scene: So now let's understand this workflow. How this workflow work. In this image, as you can see, it has a fixed custom size that the flux contact image is scale will give the output and according to the output, it will generate the image. But right now when you image rating custom node, when you use this workflow, it has a custom nodes. That is a amputee latent image. This is the extra latent space that we have generated here that will go into the latent image. But before that, the VAE encoder or the latent image was connected to this latent image. Right now we have changed the width and the height of the image and you can add it according to it. How we can do that, let's use another image and change the dimension of it. All these images that I have generated here is from Adobe Firefly. I just take the references from the Five fly. I don't have to use my brain, how many types of images I can generate. I take the reference from the firefly. You can do so. You can use Pinter firefly, anything you have, you can use or you can use Chat JPT perplexity. Is depend on you. For the ideas, I use Firefly in these lectures. Right now, let's drag this image here and let's change the width and height to 1080, 1080. 10 801-080-1080 80. This is the image and this is the pvment after the image and Latin image, we will get the result according to our squared image. Right now, let's add dragon very angry. Add dragon to the scene, very angry with fire in the Mouth. Let's run it and wait for the result. Image is not that bad, but I think image is very bad, to be honest. A red dragon to the scene very angry in the mouth flying above the sky realistic scene. Let's run it. I have changed the prompt a little bit. Let's see what type of results we get because this is really bad. So this is the result that we get right now and it is better than before because it's now flying to the sky. But yes, I don't like that much. 20. Change 1 thing at a time: This is the image, and this is the result. Results, I'm confused. What do you think? Results are good? I think results are fabulous. The way adjust the cloud, it's really awesome and just get it sketch out of it. Wow. It's awesome. There is one more thing that I want to discuss with you is that we have to understand the prom. You have to drag this image and change the lips. If you want to change the nose, you have to change the nose in the third step if you want to change the fish, color, you have to change the fish color in the fourth step. First run, you have to change the eyes, second run, you have to change the lips, third run, you have to change the nose, fourth run, you have to change the color of the fish because rating image in this context is step by step. That is why you can't just flood out the changes in this prompt and once you do that, it will not understand it or it will confuse the machine and give you a totally different result that you don't really want it. So I recommend you to rate the image one step at a time or one change at a time. Thank you for watching. See you in the upcoming lecture. 21. Add panda in the scene: Another prom that I have tried to add, add panda in the scena call talking to raccoon. Keep color style unchanged. And this is the image that I have used, and this is the image that's generated. But I like the image, but it's not panda. Let's change something, talking to raccoon. Let's change the prompt. Talking to white and black Panda or Racon in the scene, talking to White and Black Panda. Let's run it again and wait for some time. There is another prompt and this is the image that we have used, and this is the result that we have got. We tail different, but I think there might be some we have to adjust a prompt according to our need and like we have shared the link here, you can refer to this link for how to prompt for flux contexts. It's important that you do hit and trial with this and experience you get and adjusting the prompt, you will get the better result after adding lots of 22. Convert image to line art: This is the image that I'm going to use here. Let's drag it here. Let's understand the prompt first, transform this image into a cute coloring page in black and white, clean vector style, keep all elements exactly the same, including the cat force expression, backpack and scale. Mintain the glowing mountain hanging stars, clouds. These are not hanging stars, stars, clouds and surrounding nature in the same position and proportion. Remove all shading and gradients completely, use only two line widths, throughout the image, the slightly thicker outline for the main contour and a consistent thinner line for all inner details. The final results should be clean, simple, and ready for cloring. Now let's learn the prompt and wait for the result. 23. Change in any image: Use this image. And did this image to a custom size and change lips color to purple. Keep the scene unchanged. Keep the scene unchanged and keep pose unchanged. Now run it. Now we are generating another image that we took from the Adobe firefly. I took the reference from the Adobe Firefly and now changing the color of this image and make her smile. Let's see how it will look like. You can see the results are pretty fine here. It's change the lips color to purple and with Imile and change the smile. This is the result that we got here and the results are pretty fine. As you can see, we have asked it to change the lips color to purple with Imile of the woman, keep the scene unchanged and keep the pose unchanged. Everything is same, only the color of the lips and the pose of the lip that is smile is changed. I love the result, the way it adjusts the image. And one more thing we can do. Let's check another image and let's try to change it further. 24. Change in image of the eyes of subject: So right now I'm using this image that I have downloaded from Adobe Firefly. I have generated it in Adobe Fivefly from the reference. So let's open the eyes let's change unchanged and keep for same and with a smile, woman keep, open the eyes of the woman. Now let's run it and wait for the result. This is the result we get. It's beautiful. I loved it and it's really beautiful. The way open the eyes is yes, just amazing. I love it. Now I love the context even more. We will dive into further in context, influx context, and transform our images further. See you in the next lecture. Thank you for watching. 25. Convert line art into composition: Changes in the linear color. As we have already discussed about the transformation from image to linear art, let's transform this image that we have generated the linear art of into a colored image. This prompt that we are going to add here. Transform this line art of the. As you can see, this is the prom that we have generated here and let's read it out, transform this line art of the cat into a final digital painting image. Concept art style set in a lust fantasy landscape, keep the design in text, shape, same proportion and details, but render it in full color with detailed texture. Soft fabric color surrounding the cat with glowing mountains, cloud, soft mist and magical lightning. Add ambient light rays filtering throughout the cloud and tiny glowing particles in the air for an enchanting atmosphere. The overall style should be rich, colorful and painterly like a fantasy game or illustrated storybook. Let's run this prompt and wait for the result. This is the linear that we have used and I loved it. Clapping. I loved it. I loved it so much that. It's really amazing. I love it. I love it. I loved it. The way it adjusts the image. Let's compare the original image that we have already used. This is the image. It looks really nice. I can't actually say a word about it. It's fabulous. After generation, it look so nice and elegant. Everything is same. As you can see, everything is same. According to each actu, is really amazing. It is the important part that you have to change the seed to increment or decrement or randomize it. Otherwise, all the results that it will generate will be the same because of the seed value will be the same. I recommend you to change it to randomize or increment depending on you. I just use the increment so that I can follow the seed number according to the image. I love you guys because you really all my lectures and you really love my course. I really appreciate that you understand the hard work that I put in this course in creating this course, learning these new things, and sharing with you. I love that you love. Thank you for watching. I request you to please share a five star review with this and also share projects in the gallery and start discussion about the course. Thank you for watching. 26. Change into different style of images: Let's try to change this image that I have used or only dragon drop here to a Gibby art style. You can use any kind of style here. You simply have to put it that style and it will automatically change the image for you. I'm using Gibb art. We will use watercolor art. We will try to create anime art and different types of S we will use on this image. Let's see which type of output we will get according to our need. I'm using Adobe Firefly. For our references, I just try to inspire from Adobe Firefly and just generate image for you and we will try to edit or re edit that image or try to regenerate those images in our flux context, income FUI. You can use it for inspiration. As I am creative specialist, I have a subscription of Adobe Firefly, which comes with the subscription of Premiere Pro Adobe subscription. I'm using Premiere Pro after effect Photoshop. With this, we get about 1,000 arts per month. So next it will automatically get 2000. I always use for our inspiration. The way it change the image is good. Actually good. The styles are really awesome. The composition is really good. The way it's sharing the details in anime style is really good. So you can transform your images and it is really way more easy to transform your images into anything. It's actually easy to transform your images. Let's transform it to water, color, style. You can ask Chat GPT to give you the style and transform prompt, it will generate many prompts for you. You can experiment out with your images with client images with anything you want, you just experiment it out. About the generation, it took about 1.5 minute to generate images, about one, zero, 4 seconds, as you can see. This is the image that I have tried to create, as you can see. The results are pretty good, right? 27. Copy style of an image: Let's change the composition of our image to brighten the image. But yes, the way it changes the detail and keep the composition and detailing are the same as we expected. I love the results, no doubt about it. Now move to another workflow that is style references. Here is our style references. Right now I'm using this style and try to regenerate this image using our compUI style reference image. Right now, you can use any kind of size here also. These both are same files, so you don't have to worry about the file here. As you can see, it has the same latent image generator here. I'm using just here and you can experiment out with the size also, don't worry about it. Using this art style, create a new image of, uh, man. Baby, Let's try to generate and run it and see what it will do. As you can see, this is the image, and this is the image that it generated. It looked like a baby. Yes, it looked like a baby, but yes, it doesn't have a gradient. It's just bring the image into half pink and half white with the gradient of this thing. Let's try another one and see what we will get according to our apparent image. Now let's use this image and try to run this prompt using this image and use this art style, create a new image of a warrior girl with. Right now, it has bunny here. Let's change the arch style too. Fox here. Let's run it, wait for some time. Right now, the art style is good. It's actually generated the image that I have shared as a reference, but I do not want this type of here. Using this art style, create a new image of fox. Let's try this one. I created a posture of it with a warrior. Let's change it to a new image of warrior fox, not that type of image that we wanted. I have changed the prom to using this art style, create a new image of a warrior fox. Let's try the way it adjusted the warrior output is pretty awesome and pretty accurate. As you can see, on the solder as you can see, there's a spike and here is a spike, as you can see, yes, this one is different. This is extra and this is modern we have and on the forehead of the lady, according to the forehead of the lady, it has an ornament. Now, about the fox, it looks blurry. But once you enable it and run the prompt and run it will automatically resize your image and after upscale, the result will be pretty good. After this node, after this workflow, you can change it. You can use this workflow according to your desired. 28. Add logo to AI generated model cloth : So this is the image that we are going to use in our workflow right now. Let's drag it here. I'm using it in a product photography, and I have changed the prompt here. This is the prompt that I have used. As you can see, you can read it out and wait for the result. And sometime we will get the result. Let's jump to the future again. This is the result that we get here. It's look really identical. There is no change in it, but we want it on the left hand side of chest. Let's move further and change this further. Use this logo to create a product photo shoot of a shirt profess wearing the shirt is black and it has the logo on it. It has the logo on left size on left side. Of size of logo is is small. Now let's run it and wait for some time. The readers are actually good. No doubt about it. Now our product photography has done, how the logo will look like. Can you imagine the power of this com UI context? This is how we can use it in our product photography. These are some examples you can check it out and experiment it out according to you. You simply have to tell the machine that you want to use this thing and the way you want to use it, you simply have to explain it. That's all. As you can see, it look okay, but I think it should be smaller in size. I think I still have to explain it further. Like like how much in size I want it on the shirt. Can we change it further, create a product photoshoot of a military man wearing robotic suit. Professional photoshoot, and let's run it. I think I don't know what we will get, but I hope we will get something amazing. The outfit is nice. It is a military kind thing, but it is not a robotic style, but yes, the results are pretty fine. No doubt about it. This is the prong that we have used and this is the result that we have got and this is the logo that we have created, don't worry about that. Let's move to another workflow. Thank you for watching. See you again. 29. Change background of product: So this time we are using a product and place it into different environment. How we can do that. This time I have asked Chat GPT to create longer version of M prompt. Let's see, Alexa silver Wistwatch, it is not silver, it is black. W w with a burst metaltp. It's not metaltap WTP and Misticblack dial is showcased in a high end studio. Setup style like an elegant advertisement, the watch is a carefully on a rug piece of nature slate stone, partially surrounded by still crystal clear water that reflects the watch subal and adds depth. A single white orchid flower lies beside it. Its soft petals gently overlap the watch face introducing refined tropical elegance in the backlog, softly focused and exotic greenery like palm front and mona. At textures and richness without and blah, blah blah and perfume premium let's run this prom and wait for some time and jump to the future. This is the preview image that it's using. Right now we are going to use this image, which will run and we will get the results according to our image, and this is product and we are going to use a product photography. I chat DPT and with the help of hat DPT, I have created a longer version of my prom which will help me explain it further. Or sleek black luxury with a slim met finish case, minimistios code you can read it out, it's a block. You can read it out and according to it, I simply run the prompt and wait for some time. Can you guess what will the result? Whoaaa. This is the result that we got clapping for it. Amamaz amazing product photography shoot that we have done with nothing. Can you see the change? There is a change of the golden from silver to golden. I think there might be a change in the metal. Met finished, rose gold, marks, showcase gate fuul. I think there might be some change in this prom, that is why it's gold, but the results are awesome. The way it's shared the details is really good. One more thing, actually it's generating the image. I know it's not 100% perfect according to the sequel. But yes, I must appreciate I must appreciate the work that it has done to this product photoshoot. It is about 90% with the accuracy of 90%. It's really amazing. 30. Change Hair Color: So right now I have uploaded this image. This is the image that we are going to use and change the color of our image. So capoum will naturally red rich color hearing. So it should be a red, rich color and skin appear to be a realistic look visible both cinematic depth of face everything across the hair and run it. The face is same, but the hand and position is different, hair color is right. Everything is same. Let's don't use this type of and try to keep everything is same, but the hand that we have right now is different. It's totally different for our image. Let's try to change the color of hair by change everything. Change the hair of the woman to Z Position, pause, unchanged. Ce the hair of the woman to right position to t Keep pause unchanged. Facial expression. Unchanged. Let's run it. Just Wow. The weight changes, the color is really amazing. Everything is same, nothing is changed, and a simple line of fromt has changed everything. Love the results. There is no doubt about it. 31. Change Text of image: Now let's talk about change the text of our image. Right now I'm using this image and let's change the shop text to current instructor. Let's run this prompt, change the text, and I hope this work. Right now, I think it's not that good in changing the text according to our current instructor that I have put it, let's try to resize the text I'm using less character right now. I've used only current. I think it's working very fine with less amount of character. Once you increase the number of character in your image or the text, it might break it down, but at most, I think in my name, it's working fine. 32. Change the hair style of an image: So now I'm using this image. And using this hairstyle workflow and a simple prom, a woman with sleek blunt above haircut and straight banks while keeping her facial feature expression lighting overall tone unchanged. Let's earn it, and I believe it will amuse us again. Like we have talked about, this is the result. Everything is okay, but yes, there is some change. As you can see, the size of the face is little longer and here it's round shape. There might be some change in the image, but overall, the work has been done pretty. 33. Remove object from image: So to remove an object, let's go to our eight number of folder and remove items. We will use two types of remove item. We will ask context to directly remove the item by directing it to explaining the item according to our object. Let's try this one first. In, let's drag this image here and in this image, let's remove the green curs from the woman with purple hair while maintaining the same pose and scale, same lightning and everything the same. We have explained it that we have green necklace in this picture and a lady with a purple hair, we want to remove it. Let's run this prom and wait for some time. This is the image. Here's our result, as you can see, it's perfectly fine with no change in it at all, and the necklace has been removed. It's just like magic. You just ask your AI or machine learning that you want this, you write that prom, I understand your words and boom, it's give you the result. It's pretty amazing. 34. Remove object from image using inpaint: Now let's use in print to remove the same item. Let's drag this image here. And to remove in paint, you simply have to get the in print node here from the manager. If you have this in paint is missing, or we can say if it's appear in bright pink color, you simply have to go to your manager and install the missing note. Install custom nodes, missing nodes. You simply have to click here and if there is any node that is missing in the workflow, it will appear here and you have to install it. That's all you have to do nothing more. And after that, you have to restart your CFUI. That's all you have to do. Now let's move further with the neckness and removing using in print. First, we have to load the image, as you can see here, load image, right click open in mask. Now let's open in Mk area for inpnt and save. Now right click open in mask. As you can see, here is the open mask editor. You have to mask it. You can select the thickness of the brus. It depends on you. Now, it doesn't have to be perfect. You just have to mask it. That's all. Now, Save. Once you save it, the mass will appear here and you have to run the prompt again. By adding the prompt, remove the green necklace from the woman with white purple hair while maintaining the same pose scale, same lighting, and everything the same. This is the same prompt that we have used before and now run it. Wait for some time. This is the result that we got. Now let's compare all the images here. As you can see, nothing has changed. The result with the Imprint is also fabulous. Why we are using input by the way, if we want to be able to explain the object or the context or the contexts don't really understand what we are trying to say. We can use Imprint and directly ask contacts to remove this specific area or this is specific object or we can ask it to replace the object with something else. It's depend on us and depend on the situation. That's it for this lecture in the next lecture. Thank you for watching. 35. Add object into the image: So now we are going to add a silver tiara to the woman head as this is the image that we have load and we have to follow the same step like we have used before. Go to open in mask Editor, select this area using the brush. Save it run. Wait for some time, and we will get the tiara. This is the result that we wanted and it has given perfectly. Now, as you can see, we can use the power of in print and change the object, add to object, remove the object, anything we want, we can do that. Let's move further in our category, that is Laura. This is the final lecture and using Laura is really amusing. Once you load this workflow, you can see Power La loader. We can add more La here. It depends on us. Right now, we are going to use Glass beam contacts La Sapna. This is the loa that we are going to use and this is a loa that we have to download first from AI, CivitAI open source generative AI, and search for and search for context, Laura. There are many as will come up in front of you depend on the a we want it. I think the category is getting expanded here, anyone can use there are more categories are adding in the a category for context before they are not much a available here. We are going to use a specific one that is glass prism a that is prism, search for glass. Prism Here is the context tape a we are going to use. Let's click it, download this Laura in the La Download. First, you have to sign it here. After creating the ID, you simply have to search for glass, prism, click it, download this Laura. In the comp UI, you have to download in the model section search for Laura folder, and save it here. It might take some time depending on the Internet speed you have. 36. Use lora for Flux Kontext: So after downloading the La, you simply have to you can hide this and go to workflow go to edit, repress node definition. That's all you have to do. Once it's updated request, the request will update it in some time, you have to wait for some second, update, request it. Now our node definition is updated. You can check for Laura here. Here is a glass in Laura, you simply have to click it and it's selected already. By using the prompt, make this object, a glass pin with reflex and background. This is the object. Let's run it. Let's wait for some time. This is the image. Wow. Wow, wow. It's awesome. This is the same exactly screen as you can see here and this is in a grasping reflexion with a background. With this, let's conclude our course here. I'm really thankful that you have joined this course and this creative journey, and I hope that you have learned something amazing, not something. I hope you have learned a lot from this course. Remember your imagination, fuse everything you build. Keep pushing boundaries, experiment boldly, and never hesitate to remix what you have learned. A world of Art is yours to shape, embrace surprises, stay curious and let's flux context and comp be your next canvas for ideas big and small, wishing you endless inspiration and fun. See you in my upcoming course. Thank you for watching and don't forget to review this course. It really give me motivation once you review this class and don't forget to follow me. Thank you for 37. Same Characters all time : Let's talk about arc consistency. Dave contacts character consistency. As you already have noticed that we have a arrector consistency in the seventh number folder, it has included a Wblow with the name character consistency. In this workblow, I have Loda image. This is the image that I have generated within flux, and this is woman with purple hair is wearing an elegant flowing black fashion dress and striking green embroid necklace. She stand confidently with one hand on her lips, smiling and posing consistency, the background is artistically blurred with supple sunlight, casting delicate shadows, professional fashion photography high end editorial system. Once you run this prompt, let's see what we will get according to our so this is what we get. Let's compare both of these images as you can see, this is the image, this is the image, and this one is the image. As you can see, it looks with the character consistency, yes, the look is same, the hairstyle is same, a little bit different as you can see. I don't think they are a little bit different. I just put the head exactly as we want it. With the body shape structure, yes, according to a body shape structure, it's good the necklace that we have asked with a green color, yes, it's working now let's try to remove this necklace, but we will try to keep the consistency same as our arrector because we don't want the features to be changed, let's try the prompt according to that. 38. Project Details: Alright, you have hit the last lecture of this course. So let's jump into our project. I want you to generate image using flux context and fui. I want you to create a cat sitting on a window sill or a mountain landscape, so type your prompt in Cf UI and hit the run button. After that, I want you to manipulate that image using flux context. You can change the color of the or you can change the cat into a cute lion. Or it's up to you. It's your imagination, whatever you want to create, you can do so. So once you are happy with your generation, save it in your computer and head over to Project Gallery on Skillshare and share your image. That's it for this course, you have done an amazing job, and I'm really thankful that you love this course. I hope that you have learned something new in AI image generation and editing. If you enjoy this course, don't forget to follow me on Skillshare so you never miss a new class or new update, please share your project in the gallery. I love seeing your project and give a feedback on it. Also your work can inspire others in the community. Thank you for joining me. See you next time. Thank you for watching.