Home Creators Posts Import Register Favorites Logout

Downloads

Content

Hey everyone! I've just released a brand new video on Qwen Image 2.1, and this model is THE BEST LOCAL IMAGE EDITING MODEL RIGHT NOW!

Combined with my ULTRA workflow Qwen image 2.1 will allow you to do MULTI-IMAGE EDITING, INPAINTING, OUTPAINTING, FULL CHARACTER REPLACEMENT, OBJECT REMOVAL AND MOVEMENT, CONTROLNET, BACKGROUND REMOVAL and MORE!

THIS MODEL IS JUST FANTASTIC!

You can check out the new video right here: https://youtu.be/5Sby8YxbhJc

The workflow: https://www.patreon.com/aitrepreneur/posts/171329203

The installer of course automates the entire install process for comfyUI and the models downloads!

LOCAL INSTALL!!! RECOMMENDED TO START FROM SCRATCH

Download the QWEN-IMAGE-EDIT-21_ULTRA-COMFYUI-MANAGER_AUTO_INSTALL.bat

Run the bat file

Wait for the install to finish

The webui will launch. Load the normal or TURBO Qwen Image 2.1 ULTRA workflow from the workflow post, and there you go😎!

DON'T FORGET TO GO UPDATE COMFYUI BY GOING INSIDE THE UPDATE FOLDER AND RUNNING THE COMFYUI UPDATE BAT FILE THEN CHOOSING THE CORRECT LAUNCHER FOR YOUR GPU!!

IF YOU ALREADY HAVE COMFYUI INSTALLED USING A PREVIOUS INSTALLER:

Download the QWEN-IMAGE-EDIT-21_ULTRA-MODELS-NODES_INSTALL.bat and place it in the ComfyUI_windows_portable\ComfyUI folder

Run the bat file

Wait for the install to finish

???

Profit 😎

IF YOU GET AN ERROR WHEN LAUNCHING COMFYUI AFTER USING THIS INSTALLER SAYING "AssertionError: Torch not compiled with CUDA enabled" it's ok don't panic.

DM me the full error, your GPU model and which installer you used in the comments so I can help you find the correct fix for your setup.

IF YOU ARE USING RUNPOD:

The new template uses CUDA 13, supports newer GPUs like the RTX 50 series, comes with SageAttention preinstalled, automatically keeps ComfyUI updated and works with my new RunPod installers.

HOW TO USE THE NEW TEMPLATE

I made a quick video showing exactly how to deploy it and use the installers:

👉 https://youtu.be/9wlUXXyy9mE

👉 RUNPOD TEMPLATE LINK

Create an account if you haven't already: Runpod

Click on Pod (on the left side) then click deployWhen selecting your GPU on RunPod, make sure you click on Additional Filters → CUDA Versions → CUDA 13.0Then choose a GPU, click here for the comfyui template (aitrepreneur/comfyui:3.0.1-cuda13)

Then click deploy, wait for everything to finish and then click on port 8888 and then drag and drop the QWEN-IMAGE-EDIT-21_ULTRA-AUTO_INSTALL-RUNPOD.sh file on the left side of the UI then click on the "Terminal" icon on the right side on the UI

Copy and paste this line then press enter:

bash "QWEN-IMAGE-EDIT-21_ULTRA-AUTO_INSTALL-RUNPOD.sh"

Wait for everything to be installed

Then go to the port 8188

Load the normal or TURBO Qwen Image 2.1 ULTRA workflow, and there you go😎!

Also use the command:

tail -f /workspace/logs/comfyui.log

inside a terminal to get the real time logs

As always, supporting me on Patreon allows me to keep creating helpful resources like this for the community. Thank you for your support - now go have some fun😉!

Files

Comments

James Stewart

My run_nvidia_gpu.bat was running fine before I installed the auto install. Now getting this: File "C:\Users\name\Downloads\ComfyUI_windows_portable\python_embeded\Lib\site-packages\transformers\dependency_versions_check.py", line 56, in require_version_core(deps[pkg]) ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^ File "C:\Users\name\Downloads\ComfyUI_windows_portable\python_embeded\Lib\site-packages\transformers\utils\versions.py", line 116, in require_version_core return require_version(requirement, hint) File "C:\Users\name\Downloads\ComfyUI_windows_portable\python_embeded\Lib\site-packages\transformers\utils\versions.py", line 110, in require_version _compare_versions(op, got_ver, want_ver, requirement, pkg, hint) ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\name\Downloads\ComfyUI_windows_portable\python_embeded\Lib\site-packages\transformers\utils\versions.py", line 43, in _compare_versions raise ImportError( f"{requirement} is required for a normal functioning of this module, but found {pkg}=={got_ver}.{hint}" ) ImportError: tokenizers>=0.23.1,<0.24.0 is required for a normal functioning of this module, but found tokenizers==0.22.2. Try: `pip install transformers -U` or `pip install -e '.[dev]'` if you're working with git main C:\Users\name\Downloads\ComfyUI_windows_portable>echo If you see this and ComfyUI did not start try updating your Nvidia Drivers to the latest. If you get a c10.dll error you need to install vc redist that you can find: https://aka.ms/vc14/vc_redist.x64.exe If you see this and ComfyUI did not start try updating your Nvidia Drivers to the latest. If you get a c10.dll error you need to install vc redist that you can find: https://aka.ms/vc14/vc_redist.x64.exe C:\Users\name\Downloads\ComfyUI_windows_portable>pause Press any key to continue . . . I've gone to https://aka.ms/vc14/vc_redist.x64.exe and then installed that, ran again. But the same error over and over again. And now I can't use Comfyui for anything :( Please help AI Overlord!

Yukon Cornelius

I got a 3090 and images are taking like.. five minutes or more to render their edits... something wrong here or is that how long this thing takes?

Mark Lonsinger

Quick question. I have Comfy UI Desktop installed. Which installer should I use?

Mark Lonsinger

I have the 3090 also. Haven't used Qwen yet, but Krea2 takes a few minutes when I do a batch of about 4 images

tyrelli doomah

copy paste a screenshot... or copy paste cmd line text error into chatgpt plus account... tell it to take over pc and fix it for you. done. Unless you don't have chatgpt plus I guess. Just a suggestion, that's how I fix anything now. Crazy AI works and fixes anything ComfyUI.

tyrelli doomah

Text to Image was very lacking. Image editing was very awesome and fast. I got a few "linear rainbows" image creations, but I turned on 'pixel drift fixer' and it then worked creating the image. Nice model (except for text to image). Thanks

peychi

Thank goodness for the prompt encoder, as writing a decent prompt to get the result I wanted used to take me hours. I can now get rid of the Klein editor. Thank you.

Ken Kawashima

I have a 3090 and I am getting SUPER fast generations (a few seconds) with Qwen2.1 at 768x1024, like even faster than Anima. Even a batch of 4 images is still just a few seconds, so if you are getting slow generations, it's not because of the GPU, since I have the same one on a PC from 10+ years ago.

Ken Kawashima

You should use QWEN-IMAGE-EDIT-21_ULTRA-MODELS-NODES_INSTALL.bat, which only adds the models and custom nodes. HOWEVER, in my attempt it has destroyed the Python dependencies, completely breaking the VENV, so at least to me, I don't recommend it. When I tried installing it on a BLANK environment, it has worked flawlessly.

Ken Kawashima

Indeed. In my attempts, T2I generated distorted anatomy, so it's mostly useless. However, the EDITING part is great and fast, so it confirms what was said in the demo video.

Georgi Petrov

to all people with problems - there are lot of custom nodes that must be installed. Also specific models and loras. It is all in the workflow with links. I use the turbo workflow and it works fantastic with about 75 seconds for generation on 3060 gpu. The model is good at swapping and editing, with some limitations . For image generations - NO.

Aitrepreneur

Check the resolution of the first image you put it, or redownload the updated workflow, I added manual size selection.

silfras joeld

Good evening, JupyterLab isn't starting up; I've already restarted ComfyUI, but it's still the same.

Jon Force

Did the runpods break for everyone else? It gets stuck on loading Jupyter labs now and never finishes initialization. I choose the correct template and CUDA 13. Also the volume disk option seems to be gone.

Aitrepreneur

Shouldn't be, I can't test it right now since no GPU are available...but the volume disk have just been replaced by just container disk, scroll down and increase the size from 30 to 200gb

Jon Force

When it gets to the point of initializing the Jupyter, this is what the log keeps saying it just endlessly loops doing this: /start-aitrepreneur.sh: line 823: /workspace/runpod-slim/ComfyUI/.venv-cu128/bin/activate: No such file or directory

jonescrusher1

Super slow install on runpod, could be runpod issue

Alyndiar

are your Krea2 and Minimax H3 installers/workflows up to date? I'm having trouble with them

Hillsidebanger

I get an error in the image edit section: PIXEL DRIFT FIXER is missing a required input: edited_image I've just uploaded an image and tried to change the haircolor as a test. What am I missing?

Gonen Beneish

i get terrible results, even on simple prompt like: Photorealistic Slim early-30s Caucasian woman only changed prompt, nothing else, text to image.

flat

Thank you for the Qwen Image 2.1 ULTRA workflow. I run it on Ubuntu with a 5080, and it’s my daily editor. The amount of work behind these builds is exactly why I’m a paying member. I took your non-turbo workflow and, working with Claude, wired it into a terminal batch tool so I can run whole folders of images without opening the browser. Having a workflow this well organized made that practical. If it’s useful to anyone: I A/B tested int8 vs bf16 for both the diffusion model and the text encoder on several edit prompts, same seed. On a 5080 with a CUDA 13 build, int8 gave the same results at roughly a third of the time (about 33s vs 80 to 140s). The CUDA 13 note in your RunPod instructions matches what I saw. I also found faint banding at the default 25 steps that cleared at 40. Thanks again, and I’m looking forward to the next one.

Frog Prince

on 3080 generates fast enough but results are crap Task: Add 2 people from photo 2 to the field in photo 1 - Faces look completely different people - Quality at 1024 x 1024 is just blurry - With Pixel drifter off it starts just outputting the initial image without doing anything to it - With prompt enhancer on, it won't even run... This workflow is totally broken for me, but it runs without any errors?

Frog Prince

Has anyone got the prompt enhancer to work yet?

Gonen Beneish

i tried to update including dependencies, i though it might fix it, but it broke: ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. albumentationsx 2.4.13 requires albucore==0.2.20, but you have albucore 0.0.24 which is incompatible.

Frog Prince

The enhancer will crash on this setup until the underlying bug is fixed, because the cause is the setup itself. With --lowvram, ComfyUI keeps the 8.9 GB text encoder on the CPU (your log says load device: cpu), and the enhancer’s text-generation step then asks PyTorch for CUDA properties of “cpu”. Your 8 GB card can’t hold that encoder anyway, so dropping --lowvram isn’t a real fix. Your log also confirms the enhancer was on for the 20:53 run, which failed in 0.08 seconds. The other runs completed with it off. AusBoss now loads (v2.10.1, 42 nodes), so that missing-node problem is solved.

Frog Prince

How to make it work ? I habe 8gb gpu and penhancher wont work

peychi

Forked out £600 for 16GB GPU as I'm dyslexic

Frog Prince

The multi fantastic H3 media loader has a lot of issues doing what you tell it too, and a mask would be great. Don't know how to add to the fantastic loader though and your inpaint only has 1 image input thanks.

Aitrepreneur

1. Use higher res than 1024, that's too low 2. Don't use pixel drift fixer when generating a new image it's not made for that, it's made for single image edit 3. Dm me the error log for the prompt enhancer

Aitrepreneur

dm me the full error log but also yes don't use text to image, this model is more for editing than text to image gen

mnkjiu

I am having some issues with the output the images turn out okay the first few rounds i use without changing any parameters, just the prompt, the later images turns out on a lower quality, with skintone, background being somewhat smudge or blurred. any idea how to fix this?

Gonen Beneish

ah, ok, because i get terrible images when generating text to image, consider removing this section from the workflow to avoid confusion

Aitrepreneur

Can't do anything about it, it's an issue with most edit models, even chagpt or nano banana have that issue. As far as I know only the upcoming ideogram 4.5 edit model doesn't have that issue so....wait and see. For now just use inpainting if you only want to modify a special area

Aitrepreneur

As I explained in the video I put it there because people would ask for it if I didn't. Damned if I do damned if I don't lol

mnkjiu

thanks, i used your workflow and used another edit model and don't have that issue. so many it is the turbo problem.

Gonen Beneish

so how about just a comment next to it

Ash Nifty Inc

Missing Node Packs Install missing packs to use this workflow. comfyui-kjnodes 3 Installed SetNode SetNode SetNode

Ash Nifty Inc

how do you disable control net on multi edit

Gentle Sartori

Unable to start the template. When launching JupyterLab, over and over, "/start-aitrepreneur.sh: line 823: /workspace/runpod-slim/ComfyUI/.venv-cu128/bin/activate: No such file or directory"

Gentle Sartori

1 week later, this is still happening. Did I miss the fix somewhere in the comments? I increased the size from 30gb to 200 gb, but it did not help.

Aitrepreneur

dm me the full error log, have you also put cuda 13 as a filter for the pod selection?

Aitrepreneur

have you put cuda 13 as a filter for the pod selection?

Gentle Sartori (edited)

Comment edits

2026-10-16 01:00 Done. Thanks for looking.
2026-10-15 20:32 Done. Thanks for looking. I did select CUDA 13.0-13.3.

Done. Thanks for looking. I did select CUDA 13.0-13.3.

Gentle Sartori

I did. I filtered for CUDA 13.0, 13.1, 13.2, 13.3 (NOT 13.4 - that's just the way the filter came up).

jonescrusher1

Something's broken with using this sh files withtemplate on runpod, rgthree nodes messed up, can't generate anything with the workflow

Gentle Sartori

I think it was the new Global Storage container that I was trying to use. Once I switched to more traditional Network storage volume, the template worked. It looks like Global Storage [Beta] is not ready for primetime.

Seveon

I just tried yesterday and today and getting the same. I used cuda 13 as a filter.

Evgeni Damianov

Hey, guys! I have AUSBOSS node problem. Tried different versions of ComfyUI-PiD but no luck. Any ideas how to fix?

Brian Christiansen

Should i generate with fixed seed as in the video or is random seed the best Way?

Zer0 Bin0pse

any way to get this on linux? just switched away from windows and at least got base comfyui working, but now the one click batch scripts are useless

BriGuy7

is there a way to shut off the thing changing my prompt in multi edit.

Timothy Musson

In that case, download the workflow (the .json file) instead of the .bat files. Then open that .json workflow in ComfyUI. Most workflows include notes with info about required models: where to get them, where to save them, etc.