The Secret to FREE, Local AI Image Generation is Finally Here

Full tutorial link > https://www.youtube.com/watch?v=c3gEoAyL2IE

🎨 Stop struggling with complex AI image generation! In this tutorial, I reveal the ultimate, one-click solution to creating stunning, photorealistic, and stylized AI art LOCALLY on your own computer—for FREE.

🔗 LINKS & RESOURCES MENTIONED

Step 1: My Custom ComfyUI One-Click Installer

[➡️ https://www.patreon.com/posts/105023709 ]

Step 2: My Custom SwarmUI Installer & Presets Pack

[➡️ https://www.patreon.com/posts/114517862 ]

Windows Requirements Tutorial (MANDATORY for first-time setup!)

[➡️https://youtu.be/DrhUHnYfwC0 ]

Cloud GPU Tutorials (for users without a powerful local GPU)

RunPod Tutorial: [➡️ https://youtu.be/R02kPf9Y3_w ]

MassCompute Tutorial: [➡️ https://youtu.be/8cMIwS9qo4M ]

Tired of confusing workflows, endless command lines, and expensive subscriptions? Forget everything you know about Stable Diffusion and ComfyUI's complexity. I'm introducing you to SwarmUI, the revolutionary tool that leverages the power of ComfyUI's backend with an incredibly simple, user-friendly interface.

This isn't just another AI tutorial. This is a complete, all-in-one guide that takes you from ZERO to AI Art PRO in minutes. I provide a one-click installer that sets up everything you need, including pre-configured presets for achieving breathtaking realism and incredible stylization.

🔥 In This Video, You Will Discover:

The Easiest AI Art Install Ever: A step-by-step guide using my custom one-click installer for ComfyUI and SwarmUI. No coding or tech skills required!

Unlock Top-Tier Realism: Learn how to use my secret presets to generate images so realistic, you won't believe they were made on your local PC.

Master Image Editing & Upscaling: Go beyond simple generation. I'll show you how to edit images with simple text commands and upscale them to glorious 4K/8K quality.

The Secret to Speed: See how this setup optimizes performance for your specific GPU, whether you have a high-end or mid-range card.

Say Goodbye to Subscriptions: Learn how to harness the full power of state-of-the-art AI models without ever paying a monthly fee again.

This is the new standard for local AI image generation. Whether you're a complete beginner or an experienced artist, this video will change your entire workflow. It's time to unleash the full creative power of your computer.

⏰ TIMESTAMPS / CHAPTERS

• 00:00:00 Introduction to State-of-the-Art Local Image Generation

• 00:00:50 Why SwarmUI is the Best UI for Image Generation

• 00:01:28 Showcase of Incredible AI Images Generated Locally

• 00:02:22 Live Demonstration: Generating an Image in SwarmUI

• 00:03:24 Generating a Hyper-Realistic Image with the Wan 2.2 Model

• 00:04:05 How to Upscale and Add Incredible Detail to Your Images

• 00:04:44 SwarmUI's Automatic Hardware Optimizations (VRAM/RAM)

• 00:06:01 Part 1: The Ultimate One-Click Installation Guide Begins

• 00:06:43 Step 1.1: Fresh Installation of ComfyUI with Custom Installer

• 00:07:42 Step 1.2: How to Easily Update Your Existing ComfyUI

• 00:08:59 Step 2.1: Installing the SwarmUI Frontend

• 00:09:32 Step 2.2: CRITICAL SwarmUI Configuration Steps

• 00:10:11 Step 2.3: Connecting Your Powerful ComfyUI as the Backend

• 00:11:21 Advanced Setup: Using Multiple GPUs in SwarmUI

• 00:12:35 Step 2.4: Installing and Importing All The Custom Presets

• 00:13:44 Step 3.1: Downloading All Models with the Custom Downloader App

• 00:14:42 Step 3.2: Using the Image Generation & Editing Model Bundle

• 00:16:03 Model Downloader Features: High-Speed & Resume Capability

• 00:16:36 Step 3.3: How to Download Any Model from a URL (Civitai/HuggingFace)

• 00:17:48 Part 2: Using Presets & Models for Different Styles

• 00:18:01 Important Tip: Resetting Parameters Before Using Presets

• 00:18:40 Testing the Wan 2.2 Preset for Ultimate Photorealism

• 00:19:23 Testing the Flux SRPO Preset for High-Quality Realism

• 00:20:26 Upscaling the Flux SRPO Image to Maximize Detail

• 00:21:29 Creating Stylized Art with the QWEN Model Preset

• 00:22:08 Part 3: Advanced Image Editing with QWEN Image Edit Plus

• 00:22:23 Step-by-Step Guide to In-Painting and Image Editing

• 00:24:40 Image Editing Example: Changing Hair Color with a Simple Prompt

• 00:25:28 Conclusion, How to Find Your Generated Images, and Final Tips

• 00:25:51 Finding Your Saved Images and Using the Image History Tab

✨ WHAT YOU'LL LEARN IN THIS VIDEO

One-Click Installation: The fastest way to install both ComfyUI (as a powerful backend) and SwarmUI (as a user-friendly frontend).

Instant Quality with Presets: How to use my custom-made presets to achieve incredible realism and stylization without any manual configuration.

State-of-the-Art Models: Generate images with top-tier models like Wan 2.2, Flux SRPO, and QWEN.

4X Upscaling: A simple workflow to upscale your images, adding massive amounts of detail and clarity.

Automatic Optimization: Learn how SwarmUI intelligently manages your hardware (VRAM/RAM) for smooth generation, even on less powerful GPUs.

Powerful Image Editing: Use the QWEN Image Edit Plus model with simple text commands to edit images like magic.

Video Transcription

• 00:00:00 Greetings everyone. Today, I am going to show  you state-of-the-art local image generation. It  

• 00:00:07 is really hard to follow the latest developments  in image generation. Moreover, there are so many  

• 00:00:15 different information, workflows, instructions,  but today I will show you literally one click  

• 00:00:23 to install and generate the very best images  strategy. I am going to show you the latest  

• 00:00:30 state-of-the-art image generation models with  presets. When I say preset, it is really,  

• 00:00:37 really a preset. You won't be spending any  time to figure out how to make it work. I  

• 00:00:44 will show you how to generate amazing images for  realism and also for stylization with SwarmUI.

• 00:00:50 SwarmUI is so much easier to use  compared to ComfyUI. However,  

• 00:00:57 it is still leveraging all the optimizations  and features of the ComfyUI. Therefore,  

• 00:01:04 currently SwarmUI is the best interface, best  application that we can use to generate images.  

• 00:01:12 And you will be surprised that how easy  it is to install and set up. I will show  

• 00:01:18 everything. So whether you are an expert  or whether you don't know anything about  

• 00:01:23 tech and how to install these stuff,  this tutorial will show you everything.

• 00:01:28 As you can see, these images are amazing  quality. Obtaining this quality was not  

• 00:01:34 easy. I had to do a lot of research. However,  I have found the best preset for this. With  

• 00:01:41 my presets and one-click installers, you will  be able to generate these images right away.  

• 00:01:47 It will be so, so easy. Moreover,  generating this quality images is  

• 00:01:54 not requiring a very high-end GPU. All of  these images are locally generated in my  

• 00:01:59 Windows computer with SwarmUI. I mean, look  at the realism. This is an ultimate realism  

• 00:02:05 that we got locally on our Windows PC. So  this tutorial will show you everything.

• 00:02:11 If you are aiming for realism or stylization,  after watching this tutorial, you will be able  

• 00:02:16 to generate both of them locally with so much  easiness that you will be also surprised. So  

• 00:02:22 how it will work exactly? Let's make a quick  demonstration. So this is the SwarmUI interface.  

• 00:02:28 I will do Quick Tools, Reset Params to Default,  and I will do this, direct apply. Then I will type  

• 00:02:35 my prompt, whatever the prompt you want, and hit  generate. Then this will generate the base image,  

• 00:02:42 not upscaled image. Moreover, currently  it is generating 1 to 1 ratio. Let's say  

• 00:02:48 I want different ratio, for example, 16  to 9. I did set it and hit generate. The  

• 00:02:55 second one is queued. So after this one is  finished, it will generate the second image.

• 00:03:00 And you can see live the preview. So this is the  image being generated. Therefore, if you don't  

• 00:03:07 like the image, you can just click this X icon  and cancel it. But let's just wait and see. Okay,  

• 00:03:12 this is the first image. If I like this image,  next step, I will be upscaling it. Let's see the  

• 00:03:18 second image as well. And as a third image,  let's generate a realistic one. For example,  

• 00:03:24 this is the prompt I have. I am using Wan 2.2  model right now for generating these images.  

• 00:03:31 Wan 2.2 is normally a video generation model.  However, we have figured out a way, a preset,  

• 00:03:40 to generate amazing quality images with  Wan 2.2. Okay, this is the second generated  

• 00:03:46 image. Let's see the third one and then  let's upscale one of them. So you see,  

• 00:03:51 this is so easy with my presets. You will get this  interface, these presets. You will just apply and  

• 00:03:58 generate. And this is the third image. Let's say I  liked this image. So I will say Reuse Parameters,  

• 00:04:05 then in the presets, we have image upscale.  You see upscale images, direct apply, and  

• 00:04:13 generate. Now it will upscale this image into 4x  resolution. The width and height will be double,  

• 00:04:21 so it will be 4x. And we have so many amazing  presets. You see? After you followed this  

• 00:04:28 tutorial, you will have all these presets and  you will be able to use any of them as you wish.

• 00:04:33 Now it is generating the base image, then  it will upscale it and it will add a huge  

• 00:04:38 amount of details. You will be shocked by it.  And one of the best part of SwarmUI is that it  

• 00:04:44 is using ComfyUI as a backend. Therefore,  it will do automatically block swapping,  

• 00:04:51 CPU offloading, VAE tiling. So it will do all  the optimizations according to your hardware,  

• 00:04:57 like your RAM amount, like your VRAM  amount, like your GPU model. It will  

• 00:05:02 be all automatically done. 4x image upscaling  will take four times compared to generating  

• 00:05:09 base image because basically, we are  generating four times more pixels.

• 00:05:14 Upscale completed and let's look at the details.  I mean, look at this realism, these details.  

• 00:05:20 This image is just amazing. When we compare to  the base image, this was the image resolution,  

• 00:05:27 the quality, the details, and this is the new  image. You may notice that the image changed  

• 00:05:33 slightly more than what we expect. We can prevent  it. You can reduce the refiner control percentage,  

• 00:05:40 which will be automatically enabled in here. So  if you make this lesser, like 50 percentage, it  

• 00:05:46 will change image lesser degree. It is up to you.  So based on your needs, you can reduce this and  

• 00:05:53 it will make it more like the base image, maybe  a little bit lesser degree of adding new details.

• 00:06:01 So how easy and how to install SwarmUI, these  presets, ComfyUI backend? So easy. First of all,  

• 00:06:10 we are going to install ComfyUI. You will  find this link in the description of the  

• 00:06:15 video and also in the comment section of the  video. Always download the latest zip file.  

• 00:06:21 And as you scroll down, you will see the  change logs. You can read them to learn,  

• 00:06:26 but one thing is important, you need  to follow Windows requirements tutorial  

• 00:06:32 before starting installation. This is one  time mandatory and after that you will be  

• 00:06:37 able to run all the AI applications without  following this again. This is just one time.

• 00:06:43 First, I will show a fresh installation. So I  will install it into my Q drive and right click  

• 00:06:49 and extract all like this. Make sure that  your folder name doesn't have any special  

• 00:06:55 characters or any space characters. All I need  to do is just Windows install or update ComfyUI  

• 00:07:01 and run. And then according to your Python  version, select it. I still recommend Python  

• 00:07:07 3.10.11. So I will choose the option 1, but we  support all these. My installer will install  

• 00:07:14 it with SAGE attention, FlashAttention, Triton,  xFormers. Everything is pre-compiled. Moreover,  

• 00:07:21 it will automatically install ComfyUI  Manager, ComfyUI IPAdapter, ComfyUI GGUF,  

• 00:07:27 and Res4LYF. Furthermore, if you need Reactor or  Impact Pack extensions, you can just run their bat  

• 00:07:37 files after installation has been completed,  but you don't need it for this tutorial.

• 00:07:42 So what if you need to update your existing  ComfyUI? But be careful that you had to have  

• 00:07:48 installed my ComfyUI, not other installations.  Enter inside your previous ComfyUI folder,  

• 00:07:55 paste the zip file, make sure to quit from  your running application if you have any,  

• 00:08:02 then right click and extract  all. I am using WinRAR,  

• 00:08:05 you can use anything. Overwrite all the  files that are updated. Then double click  

• 00:08:10 Windows install or update ComfyUI and run.  When you're updating previous installation,  

• 00:08:16 it will be really fast. Just follow the speed  here. You see, this is updating right now. Okay,  

• 00:08:22 updating. It should be done under 1 minute  in total time. Okay, still updating,  

• 00:08:28 almost done. We are getting there. Yes, updated.  It took like 30 seconds to update the ComfyUI to  

• 00:08:35 the latest version with the newest stuff that we  need. This is how my ComfyUI installation works.

• 00:08:42 Okay, the fresh installation also completed.  It took few minutes. You can scroll up to see  

• 00:08:48 whether there are any errors or not. There  shouldn't be any errors. Once you see that,  

• 00:08:53 press any key to continue. You are done with the  ComfyUI installation. As a next step, we are going  

• 00:08:59 to install SwarmUI. The link of this post will be  also in the description of the video and also in  

• 00:09:05 the comment section, the pinned comment. Download  the latest SwarmUI model downloader zip file. Move  

• 00:09:11 it into the disk wherever you want to install.  So I will install into my Q drive. Extract it,  

• 00:09:18 then you will see that there is Windows install  SwarmUI.bat file. Double click, run. This  

• 00:09:24 will install the latest version of the SwarmUI.  SwarmUI installation is also lightning fast. Okay,  

• 00:09:32 it is getting there. Yes, started. Now follow  my steps exactly. I agree. Customize settings,  

• 00:09:40 not just install. Yes, we can say yes. Choose your  theme. Okay, next, next. Now this is important. We  

• 00:09:48 are not going to use ComfyUI local. Okay, this is  super important because this ComfyUI version is  

• 00:09:54 very weak and problematic. We are going to  use the ComfyUI we just installed. This is  

• 00:09:59 mandatory. So select none, next. You can download  the models that you want. I'm not going to. Next,  

• 00:10:05 and yes, I am sure. You see, it is instant,  instantly installed. Then we are going to go  

• 00:10:11 to server tab, backends, and we are going to  add ComfyUI self-starting. Okay. You see now  

• 00:10:19 this screen has arrived. We are going  to add our ComfyUI folder path here.

• 00:10:25 So I had installed it into my Q drive, which  is inside here. ComfyUI, copy this path,  

• 00:10:32 and then backslash main.py. So then you can also  give some extra arguments. What kind of extra  

• 00:10:40 arguments? For example, in the ComfyUI post,  you will see that there is like reserve VRAM,  

• 00:10:46 use SAGE attention, GPU only. This is for  very high VRAM GPUs. High VRAM. There are  

• 00:10:52 more options of ComfyUI, but I am going  to use SAGE attention. The rest can be  

• 00:10:58 default. This SAGE attention will improve my  performance significantly. Save. In some models,  

• 00:11:07 this can produce black output. If you get black  output, just edit here and delete this. Otherwise,  

• 00:11:15 you can keep it. And currently it is going to  use my first GPU. If you have multiple GPUs,  

• 00:11:21 you can add another backend like this. Copy this,  paste, then select the GPU ID 1. You can use extra  

• 00:11:30 arguments, save. So now when I generate two  images, it will generate on each GPU. Moreover,  

• 00:11:37 I also recommend overQueue to zero. So it  will immediately start on the next available  

• 00:11:44 GPU without waiting. So these are the settings  that I am using. I don't change anything else.

• 00:11:50 And when I start the SwarmUI again next time,  it will also automatically update my ComfyUI.  

• 00:11:55 You can see that there is always auto-update of  the backend. If you get update errors, which you  

• 00:12:01 shouldn't, you can manually update the ComfyUI  because we have updater. You see, we just run  

• 00:12:07 this. For SwarmUI update, we have Windows update  SwarmUI. Let me show you that too. So, okay, it is  

• 00:12:14 started, so I will just close this. Double click  Windows update SwarmUI.bat file. It will start  

• 00:12:20 right away. You see, it is starting, and yes.  Now I need to wait again for backends to load.  

• 00:12:27 So currently, in my installation, there is no  presets here. How am I going to use the presets?

• 00:12:35 We have preset installer. You can use both  ways. When you use the preset installer here,  

• 00:12:41 windows_preset_delete_import.bat file, it  will delete your older presets and import  

• 00:12:47 new ones. But don't worry, your older presets  will be saved in preset backup. Currently,  

• 00:12:54 since I don't have any presets, they are  not saved anywhere. Then I need to click  

• 00:13:00 refresh and all presets arrived. So  now you are ready to use like this.

• 00:13:06 And if you want to update your previous SwarmUI  installation instead of a fresh installation,  

• 00:13:11 go to the folder, extract zip file, right click  and extract all files here. You can use Windows  

• 00:13:18 extract system or whatever. Overwrite  all files. Then it is same, you just  

• 00:13:23 need to import new presets. You can import new  presets from here, import presets, choose file.  

• 00:13:31 So this is manual way. Choose the preset  file, overwrite existing presets and import,  

• 00:13:37 and it will import existing presets for you. Then  click refresh. We have made a fresh installation,  

• 00:13:44 but we are lacking the models. When I  go to models, there are no models here,  

• 00:13:49 or there are no LoRAs here. How you are going to  find these models? I made it so easy as well. So  

• 00:13:55 this was our fresh installation and you see there  is windows_start_model_downloader_app.bat file.  

• 00:14:02 Double click it, click run. It will start our  model downloader application. This application  

• 00:14:08 is amazing, you will see it. It will also generate  a virtual environment. All my installers generates  

• 00:14:14 a virtual environment and installs everything  there. So your installed ComfyUI will never  

• 00:14:20 break anything else. You can have multiple ComfyUI  installations. All will be independent. You see  

• 00:14:25 when I enter inside ComfyUI, there is virtual  environment. So everything is installed here,  

• 00:14:30 nowhere else. So I can have as many  as different ComfyUIs that I want.

• 00:14:34 Okay, it is getting started. Yes. So this is our  model downloader interface. For this tutorial, we  

• 00:14:42 are going to use SwarmUI bundles, Image Generation  and Editing bundle. This will have QWEN Image  

• 00:14:48 Edit Plus, QWEN Image, best upscaler models, face  segmentation models, Wan 2.2 text-to-video model.  

• 00:14:57 This is what we are going to use. Flux SRPO model,  this is best Flux realism model, and the other  

• 00:15:05 needed files. So just click this and follow the  CMD window. This downloader is extremely robust.  

• 00:15:13 It will download and do hash calculation.  Therefore, all the downloaded models will  

• 00:15:19 be verified, 100% accurate. There can be never  any corrupted downloaded model file with this  

• 00:15:26 downloader. The logic of this downloader is that  it is like you get. If you have used you-get, you  

• 00:15:33 will know it. So when I go back to my models  folder into diffusion models, this is usually  

• 00:15:38 where the models downloaded. You see that it is  16 parts. So it is downloading with 16 different  

• 00:15:45 connections. Therefore, it is super fast. You see,  my speed is over 100 megabytes per second on my  

• 00:15:52 personal computer. On cloud machines, this is 300,  400 megabytes. When I go to Task Manager and my  

• 00:15:58 internet connection, you can see that it is using  1 gigabit. This is my internet speed. Therefore,  

• 00:16:03 it is maximum utilizing it. Once these model  downloads completed, it will merge these  

• 00:16:10 files into a single file and it has a full resume  capability. So I will close this to show you. Then  

• 00:16:17 start the Windows Start Models Downloader file  like this. Then I will click the SwarmUI bundles,  

• 00:16:24 download image generation and editing bundle. And  you see, it is continuing wherever it is left.

• 00:16:30 Believe me, this application took a lot of time  of me to develop and this is version 85 right now.  

• 00:16:36 So when you go to URL downloader, you can also  download models from this interface. For example,  

• 00:16:43 let's download this Flux model. Copy this URL,  then paste the model URL here, and select the  

• 00:16:50 target folder. Since this is Flux model, let's  see, this is a Flux pruned fp8 model. So this  

• 00:16:57 goes into models, diffusion models. Then click  download. It will automatically start downloading  

• 00:17:03 this model as well. However, currently we are  downloading other models, so it will show both  

• 00:17:09 of them at the same time, but it is so easy.  You can also provide your own Civitai key or  

• 00:17:16 Hugging Face API key, but we have some default  keys, so you don't need to provide them. So  

• 00:17:22 this URL downloader is also extremely useful,  especially on cloud machines, on cloud services.

• 00:17:28 So you see this model file has been downloaded,  QWEN Image Edit Plus, and it has been merged.  

• 00:17:34 Now it is downloading the other files. And  when you next time start the application,  

• 00:17:39 the verified downloaded files will be just  skipped. It will not re-download them,  

• 00:17:45 or if the download has been  interrupted, it will resume.

• 00:17:48 And the rest is as I have shown in the beginning.  I have returned back to my other installation  

• 00:17:54 where I have all the downloaded model files, as  you can see right now. Before using any preset,  

• 00:18:01 make sure that click Quick Tools, Reset Params to  Default. So it will default all the parameters.  

• 00:18:08 Then go to preset and use whatever you want. For  example, let's use the QWEN Image Realism. This  

• 00:18:15 has significantly improved my quality with  QWEN LoRA. I will show hopefully soon it,  

• 00:18:21 training QWEN LoRA. Direct apply and paste  your prompt. Then select the aspect ratio  

• 00:18:27 that you want. This is also really fast. It  is only eight steps, but normally I recommend  

• 00:18:33 QWEN for stylized images. So for stylized  images, you can use this preset rather than  

• 00:18:40 this one. For realism currently I recommend Wan  2.2 image realism preset, and for best quality,  

• 00:18:47 I recommend you to upscale images to 2x. Actually,  it is 4x resolution in total. So we are getting  

• 00:18:54 the QWEN image result. I will also show you  Flux SRPO. It is also a really realistic model.

• 00:19:02 So let's see the results of each one. You see  the QWEN is really fast and this is a really  

• 00:19:07 high resolution, it is almost 2 megapixels.  Okay, the image has been generated. It is not  

• 00:19:11 very realistic. The QWEN is not very realistic.  This is the realism we got, but this is what QWEN  

• 00:19:18 is. Hopefully, you will see that it is way more  realistic with self LoRA training. So let's try  

• 00:19:23 this with Flux SRPO. Quick Tools, Reset Params to  Default, then direct apply. This will select the  

• 00:19:31 Flux base model by default. Therefore, you need  to select Flux SRPO yourself like this. You see  

• 00:19:37 the selected model is now Flux SRPO. Paste your  prompt, select your aspect ratio, okay, here,  

• 00:19:44 and generate. Flux SRPO is also really fast to  generate. Now it is reloading my model. One of  

• 00:19:50 the best part of my preset system is that you can  test all of them so quickly, compare, and decide  

• 00:19:58 which one is working best for your case. If you  remember, this image was really, really realistic  

• 00:20:03 on Wan 2.2 preset. So for realism, currently Wan  2.2 image realism is my number one choice, and  

• 00:20:12 Flux SRPO is my second choice, which we are about  to see. And this is not upscaled, remember we need  

• 00:20:19 to upscale for even better. Yes, this is Flux  SRPO. It is also really realistic. If we upscale  

• 00:20:26 this image, it will become much better. So reuse  parameters, direct apply the upscale. You will  

• 00:20:31 notice that it has enabled upscale and generate.  Now we will get much better quality image.

• 00:20:38 If you also don't have a powerful GPU and you want  to use cloud services like RunPod or MassCompute,  

• 00:20:44 I will make separate tutorials to show how to  install and use there. Moreover, I already also  

• 00:20:51 have some tutorials for RunPod and MassCompute.  For example, this tutorial is made for RunPod,  

• 00:20:57 so you can watch this to learn how to install  ComfyUI and SwarmUI. And this tutorial is showing  

• 00:21:03 you how to install new ComfyUI on MassCompute  and use the existing SwarmUI there. So we already  

• 00:21:11 have tutorials, but I plan to make specific  tutorials for this one as well, for today's  

• 00:21:17 tutorial as well. Okay, now I have upscaled  the image. So from this image to this image,  

• 00:21:23 we got it. So it is up to you, you can try, see  which one is working best for your case. And for  

• 00:21:29 stylization, I think QWEN is the best. So I will  do reset params to default, direct apply this one,  

• 00:21:38 and let's try a non-realistic image. So let's  make the resolution like this. Surprisingly, Wan  

• 00:21:44 2.2 also good with non-realistic images as well.  For example, this one was generated on Wan 2.2,  

• 00:21:52 or this image was also generated on Wan 2.2.  But in my opinion, if your aim is not realism,  

• 00:22:00 then QWEN may yield better results. This is  base one, not upscaled, and this is the quality.

• 00:22:08 So what about image editing? We have amazing  preset for image editing as well. Reset params  

• 00:22:13 to default. And currently, best image editing  model is QWEN Image Edit Plus 12 steps. So direct  

• 00:22:23 apply. And I am going to edit these two images. So  I will drag and drop them to here. You see, here,  

• 00:22:30 not init image. This is important. Then select  the aspect ratio that is closer to your input  

• 00:22:38 main target image. If you don't know it, you  can use this strategy. Choose your main image,  

• 00:22:43 click res, and use closest aspect ratio like  this, and it will auto-set. Or you can use exact  

• 00:22:49 aspect ratio as well, or you can use resolution  as well. So you can compare all three cases and  

• 00:22:55 decide which one is working best. And after that,  disable init image. This is important. Then type  

• 00:23:01 your prompt. So my prompt will be simple as this  one. Make him wear headphones and generate. This  

• 00:23:10 QWEN Image Edit model works with commands. So  you give a command, describe whatever you want,  

• 00:23:19 and just wait the result. This model is really,  really powerful. I already have full tutorial  

• 00:23:24 for this model. You can see it here, Nano Banana  versus QWEN Image Edit or QWEN Image Edit full  

• 00:23:32 tutorial. We also have a newer model, QWEN  Image Edit Plus. I already updated the model  

• 00:23:37 downloader and also the preset for this new  model. So even though these older tutorials  

• 00:23:43 were for the base QWEN Image Edit model, now  we have the latest QWEN Image Edit model.

• 00:23:49 Okay, it didn't understand the prompt. So let's  try another prompt. Make him wear the headphone.  

• 00:23:55 This usually never happened to me before, but  this time it happened. So let's try again. Okay,  

• 00:24:00 this time it is working. So if you don't  get the output you want at the first time,  

• 00:24:06 try with a different seed. It will help you.  You see, I am currently trying a different  

• 00:24:11 seed. One of the best parts of the SwarmUI is  that you can have different presets yourself,  

• 00:24:18 you can have wildcards to test different  prompts. For example, I have used a lot of  

• 00:24:23 different prompts to generate to show you in  the beginning. These prompts are also shared  

• 00:24:28 in the zip file. You will see that there  are some test prompts. So you can follow  

• 00:24:34 these prompts and see which one of them, how  written they are. Okay, this is it. You see,  

• 00:24:40 we made him wear that specific headphones.  This is the best image editing model. I mean,  

• 00:24:45 you can make like this, for example, make  hair blue and generate. It understands your  

• 00:24:54 instructions. It is like ChatGPT image editing  or Nano Banana image editing. And this model is  

• 00:25:00 really, really good. If you don't get exact image  that you want, I mean like pixel distortions or  

• 00:25:07 some other stuff, you should try different aspect  ratios from here. You see, closest aspect ratio,  

• 00:25:13 exact aspect ratio, use resolution, use exact  aspect resolution. So try all the cases and see  

• 00:25:20 which one is yielding the best results. And this  is it. So it did the command I made him to do.

• 00:25:28 So this is all for today. I hope you have enjoyed.  Please subscribe, leave a like, leave a comment.  

• 00:25:34 The links will be in the description of the  video. And hopefully I will make quick RunPod  

• 00:25:39 and MassCompute tutorials as well to show you  how to install on cloud. And if you don't have a  

• 00:25:45 powerful GPU, you will be able to generate really  fast images there. Moreover, all your images are  

• 00:25:51 saved inside SwarmUI, inside output, inside local,  inside raw, based on date. If you have generated  

• 00:25:58 grids, based on grid. Furthermore, in the image  history tab, you will see all your images. You  

• 00:26:04 can filter them by words. It really works well.  Like eyeglasses, and it will show all the prompts  

• 00:26:12 that had eyeglasses. And you can look each one of  them like this one. And hopefully, see you later.