Post "1-Click INSTALL QWEN IMAGE EDIT 2509 (PLUS) ComfyUI WebUI!" by "Aitrepreneur" from patreon

LW

I opened ostris AI-toolkit but I didn't see an option for training loras for Klein. Is this something in the works? Or is a different lora trainer needed?

RavenWT

Just reinstalled the workflow, and strangely we have a lot of errors to fix, and when running it, it doesn't modify the image animore or just lightly.

RavenWT

Got the same erreur, and you need to go inside : \ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-GGUF Then search for Int32 and the line: qs = torch.gather(kvalues, dim=-1, index=qs.to(torch.int32)).reshape((n_blocks, -1, 32)) There you need to switch the in32 for int64. Your line is supposed to look like this: qs = torch.gather(kvalues, dim=-1, index=qs.to(torch.int64)).reshape((n_blocks, -1, 32)) It fixed my issue

Michael Ang

Hi I keep getting the error in Comfy UI - TextEncodeQwenImageEditPlus gather(): Expected dtype int64 for index

a pretty parrot

i believe fixed this by swapping the text encoder, you need one that matches the qwen model your loading.

westingtyler

any time I do anything i get this error: mat1 and mat2 shapes cannot be multiplied (748x1280 and 3840x1280) i have an rtx 5070ti 16gb, using win 10. any idea how to fix this? I use all the same resolution of image, but I always get this.

David Murakami

""AssertionError: Torch not compiled with CUDA enabled" it's ok don't panic." Sorry, but I'm still panicking! Like lots of people here it seems, this is not working despite the torch reinstall suggestion. Please help, thank you!

Matthew Butcher

Could you please post a video on setting up a fresh install. I think there are alot of assumptions as to python versions, path variables etc. It would be good to start from scratch.

MB

Hello. I am running the installer QWEN-IMAGE-EDIT-PLUS-COMFYUI-MANAGER_AUTO_INSTALL.bat with my pc that has an nvidia 5090. Then I run the update file QWEN-IMAGE-EDIT-PLUS-MODELS-NODES_INSTALL.bat I receive the message below despite I have done these steps I went inside the python embedded folder, click on the folder path, type cmd press enter, and typed: python.exe -m pip uninstall torch python.exe -s -m pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu130 The error report is below Please help C:\Users\marco\Downloads\ComfyUI_windows_portable>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build [START] Security scan [DONE] Security scan ## ComfyUI-Manager: installing dependencies done. ** ComfyUI startup time: 2025-10-18 12:52:52.512 ** Platform: Windows ** Python version: 3.12.7 (tags/v3.12.7:0b05ead, Oct 1 2024, 03:06:41) [MSC v.1941 64 bit (AMD64)] ** Python executable: C:\Users\marco\Downloads\ComfyUI_windows_portable\python_embeded\python.exe ** ComfyUI Path: C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI ** ComfyUI Base Folder Path: C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI ** User directory: C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\user ** ComfyUI-Manager config path: C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\config.ini ** Log path: C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\user\comfyui.log Prestartup times for custom nodes: 0.0 seconds: C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use 0.0 seconds: C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy 1.9 seconds: C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager C:\Users\marco\Downloads\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\cuda\__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. import pynvml # type: ignore[import] Checkpoint files will always be loaded safely. Traceback (most recent call last): File "C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\main.py", line 149, in import execution File "C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\execution.py", line 15, in import comfy.model_management File "C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\comfy\model_management.py", line 237, in total_vram = get_total_memory(get_torch_device()) / (1024 * 1024) ^^^^^^^^^^^^^^^^^^ File "C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\comfy\model_management.py", line 187, in get_torch_device return torch.device(torch.cuda.current_device()) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\marco\Downloads\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\cuda\__init__.py", line 1069, in current_device _lazy_init() File "C:\Users\marco\Downloads\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\cuda\__init__.py", line 403, in _lazy_init raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled

Mailman

It should be possible to swap a person (image 1) with a person in image 2, shouldn't it? I can't get it to work. the overall image 2 should be the same, just with 1 person swapped in/out.

Tokyojunkie

Wish we can do a batch output from video. Maybe we can combine with ebsynth?

Emilien-Edmond D'aramitz (LeonceTheLion)

Just want to put this here brothers and sisters. I've noticed that with this and the previous runpod installer for the non 2509, that sometimes, but not all the time it changes the port of your comfy. I highly doubt this is an issue with Aitrepreneur's code, but rather a glitch with runpod. If you notice your comfy isn't loading, or is spinning, look at your comfy log, and you may notice it is 3001. In my case, i was using the ultimate SD workflow, because it comes with automatic1111 and an uploader, so i can download civitai models, so it is 3020 for me and tonight it changed the port to 3021. The solution is very easy, edit your pod, and open up that port, add the new port to the list. You do not need to delete the other port, just open the new port. **Edit** You may have to rerun the script if for whatever reason runpod wipes your workspace afterword, as it unexpectedly did to me tonight, just go into your models and see if they are there, or if you notice your comfy reverted to a previous version.**, Sometimes it comfy manager disapears with this workflow, so its just best to use Aitrepreneur's comfy workflow.... we need a civitai downloader for comfy, the one i found i couldnt get to work because i am an novice.

Emiliano

Hey I've installed from scratch 3 or 4 times and keep getting this error, if anyone has encountered this and can help I really appreciate it: Edit: Local install, 5070ti w/ 16gb vram. I'm selecting the 12-16gb option in the installer. TextEncodeQwenImageEditPlus CUDA error: no kernel image is available for execution on the device CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions. In the terminal: RuntimeError: CUDA error: no kernel image is available for execution on the device CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

Daniel Liu

hello, i ran this and the image coming out is blurry, can you advice what is wrong?

mnkjiu

Just sharing, doing Qwen Image Edit Plus on macOS is not possible. Even with the mmproj file. Unless someone can share with me how to do it properly with the right files

Peter Noble

I installed using your batch file: QWEN-IMAGE-EDIT-PLUS-COMFYUI-MANAGER_AUTO_INSTALL.bat Then I updated comfyui to the latest version because it said some nodes needed newer version. Now I get this error message: UnetLoaderGGUF cannot reshape array of size 1618850 into shape (3072,2112)

Lex Jong

Hey Aitrepreneur, I run a 5070ti and used your installer and ran into the AssertionError: Torch error. But everything worked after the workaround (which torch) you placed in your guide. But now I have some other issues, which I can't seem to fix (tried all weekend), can you take a look and give some advice? With Qwen-Image-Edit-2509-Q5_K_S.gguf in Unetloader and one of the following in CLIPloader: + Qwen2.5-VL-7B-instruct-UD-Q4_K_S.gguf: I get Falling back to numpy dequant for qtype: 23 (which is a fallback to CPU I understand. -> which does finish but very slowly + Qwen2.5-VL-7B-instruct-mmproj-BF.gguf: I get the error TextEncodeQwenImageEditPlus - mat1 and mat2 shapes cannot be multiplied -> which stops the proces + Qwen2.5-VL-7B-instruct-Q4_K_M.gguf: I get the error CLIPLoaderGGUF - Unexpected text model architecture type in GGUF file: 'qwen2'

Anthonin Hauy

Hey ! everything works well but I have ghosting problems, gen are very blurry.

Cristo

This a great tool but unfortunatelly after the update you made, the template, workflow is waaaay slower to load and to the generate images, even if I want to mask something, it takes more than 1 minute. Hopefully you fix it. I'm using it on runpod. I've restarted the pod many times, I even deleted the storage and created a new one, I download everything again and still face this same issue

PinkE

I'm using your run pod template. I updated Comfy UI, but after refreshing the page there is no longer a Manager button anywhere. Without the Comfy UI manger and it's restart button, I don't know how to restart the server. What can I do?

Aitrepreneur

you can try to uninstall the controlnet aux node and reinstall it

Anthony Sparks

OpenposePreprocessor No module named 'dateutil im getting the above error i have updated and done what was in the video is there a way i can fix this?? its only when open pose is trying to be used!

Aitrepreneur

DM me screenshots

Anthony Harris

I did a clean install of this and downloaded the new workflow but I am still getting the ghosting effect in the image output of the altered image

Aitrepreneur

Yup, that's my bad, you can redownload the installers now, it's fixed

Aitrepreneur

you probably didn't open the terminal in the correct folder when putting the installer

Aitrepreneur

have you used the installer? and do you have the Qwen2.5-VL-7B-Instruct-mmproj-BF16.gguf file in the ComfyUI_windows_portable\ComfyUI\models\text_encoders folder?

Aitrepreneur

Yup, my bad, you can redownload the installers now, it's fixed

Cristo

After the updated, I can't auto install it on rundpod. It says "there is no such file"

westingtyler

hello! I'm using the QWEN IMAGE EDIT PLUS ULTRA in comfyui, which I manually updated, got models and nodes. trying the top left section - no inpainting, no controlnet, when I get to the Switch latent, whether it's on true or false, I get this: mat1 and mat2 shapes cannot be multiplied (784x1280 and 3840x1280). I'm using a 1024 image in the top slot, and the width and heighth are their default, 1024x1024. RTX 5070ti 16gb, win 10. using Qwen Image Edit 2509 Q5 K S .guff and the other default models of the workflow.

bigsk1

you need huggingface token if downloading via cli, or download in browser and move to folder

Garrulous Skeptic

Unauthorized for url: https://huggingface.co/lllyasviel/Annotators/resolve/main/body_pose_model.pth (Request ID: Root=1-68de244a-7c09185314e732ec6b0f9b80;0e1b4c59-b320-4f5a-8895-0ca92fcab5db) Invalid credentials in Authorization header

Christian LOUIS

I tried the runpod version but i get this error • downloading Qwen-Image-Edit-2509-Q8_0.gguf ####################################################################################################################################################################################################################### 100.0%####################################################################################################################################################################################################################### 100.0% • downloading Qwen-Image-Edit-Lightning-8steps-V1.1.safetensors curl: (22) The requested URL returned error: 404

Foxy

I'm getting the response "No such file or directory" after I drop the two lines of code into the terminal. Any ideas?

Erock Vaughn

civitai dot com you can find many there.

Robb

Where do I find the NSFW Loras for this that you were taking about?

Cristo

I noticed that the images that I change position or clothes, they get a lot of contrast

Jeff

Hi! I'm getting the following error when trying to run Qwen plus: "UnetLoaderGGUF This model is not currently supported - (Model architecture not allowed for conversion! (i.e. reference VS diffusers format))" any help would be fantastic!! Thanks! (I'm running the nightly version, with comfyUI V3.37, and Qwen-image-Edit-2509-Q5_K_S.gguf ) should I try using Q5_k_M? or maybe go up or down to Q4 or Q6

Aitrepreneur

read the post, it explains how to solve this

Aitrepreneur

dm me screens

Aitrepreneur

yes

Aitrepreneur

could be bad prompts or incorrect resolutions

Aitrepreneur

I could but it wouldn't be interesting since it's the same as training the kontext model

Aitrepreneur

read the post, you can easily solve the error, this often happens because of the custom nodes updates, I'll try to fix this next time

Aitrepreneur

You need to use the installer and have the Qwen2.5-VL-7B-Instruct-mmproj file inside the unet folde

Aitrepreneur

because you ran the installer as admin, you can just move the folder somewhere else since it's a portable version

Flo_Tou

how did you do it ?

Sean Wilton

Hey I hope someone can help me, when I try to run a script in ComfyUI I get the following error, any help would be great, Thanks! CUDA error: no kernel image is available for execution on the device CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

Tommario Minor

I had to manually bypass the True>openpose false >depth node. disabling the controlnet fast groups bypass does not disable this node and it is throwing an error when not using controlnet

Tommario Minor

Prompt execution failed Prompt outputs failed validation: Switch any [Crystools]: - Required input is missing: on_true - Required input is missing: on_false I'm getting the same errors, i have painstakingly deleted my complete comfy install and did a fresh install for the 4th time. I've ran the update.bat after the fresh install. I still encounter the same errors. please help, I appreciate what you do but i cant keep paying for this every month if i can not make use of the tool.

Tommario Minor

same here. does any one know where to get the more simplistic workflow that does not incorporate all of these open pose nodes. ive disabled them and still get this same error.

Extreme Mojo

I've been using the origional Qwen loras that I trained. They work Awesome.

Extreme Mojo

This is absolutly Awesome! Is there a way to use DW pose? And when the heck are the VACE people gonna get the WAN 2.2 thing going... It's a very exciting time for Ai Creators.

phail216

You need update comfyui from the update folder

mnkjiu

No issues with the install. Working fine. Are there any LORAs we can use?

Alex Spin

How can I intall all this missing files/nodes ? ReferenceLatent CFGNorm TextEncodeQwenImageEditPlus ResizeMask GrowMaskWithBlur ImageResizeKJv2 PrimitiveStringMultiline That's it shows after running the COMFYUI MANAGER AUTO INSTALL BAT Same happenwith other pre installed COMFYUI that I had, it was updated and I've used the MODELS-NODES_INSTALL BAT ... After that, Same missing files/nodes

FanArtFuseBeads

A algunos mecenas se nos olvida simplemente agradecer tu trabajo por facilitarnos la vida en el filo de la tecnología cuando no tenemos tiempo ni interés por estar todo el tiempo actualizados. GRACIAS.

bigsk1

your at the very beginning of the comfyUI journey, I suggest watching many youtube videos to get the basics down.

Vincent Nestler

I am on a windows laptop with a 5090. I have the comfyui.exe installed from https://www.comfy.org/ where do i run model, nodes install from?

sai ram

Make wan2.2 animate video @aitrepreneur

Ivan L

Yup same problem, when i do a 1 image edit (and turning off all the controlnet stuff, results in a ghosted image on image blurry effect. Regardless if I try different weights q8,6 or lighting 4step/8step. etc.

kristopher ulley

i have a 4090 and tryed bypassing this in the q8 version and it stalls, in the q5 version it creates a blurry photo with bad contrast.

kristopher ulley

Prompt execution failed Prompt outputs failed validation: Switch any [Crystools]: - Required input is missing: on_false - Required input is missing: on_true

kristopher ulley

I got this issue as well. Did you find afix?

Bowen Baars

Will you also make an install for wan 2.2 animate?

Daniel

Mostly works. Just two weird parts, When using just one image like you did in the video on the three image workflow I get a error: Prompt outputs failed validation: Switch any [Crystools]: - Required input is missing: on_false - Required input is missing: on_true And when trying to use more then one sometimes the images try to blend instead of doing what I ask. Like change style of image 1 to that of image 2, then I get a bunch of people and elements that blend into the people in image 2. But maybe it just not as good at that yet. Anyone had similar issue? Thanks anyway its great to try it out. Hope we get normal qwen to with the controlnets.

bwild370

I love this, Could you make a short video on how to run Ai Tool Kit to train a Qwen image Edit Plus Lora? It would be an awesome video??? :D

Dragonbait

hmm, issues with this one. My images are coming out like weird double images, almost like 2 images slightly off from each other !?

Dave Spicer

Apparently the Q5 model downloaded by the batch file was corrupted - another user said getting from Huggingface fixed the issue. Our overlord said he'd download it again so the batch file would use an uncorrupted copy.

Dennis Stolte

hi Aitrepreneur, thx for making there bat files. i just wish you would include one just for getting the models. This is the second time i used one of your installer bat files on a perfectly working comfyui installation and it just bricked it. while all i needed were the models. Last time i was able to get it fixed again with a backup this time however i did not make a backup. I started the bat file, choose the type of file for my vram that was asked and i went.. "shit! not again." uninstalling and reinstalling all kinds packages ruining my installation. Should have just found the models. Me sad now

Ronald Tang

Im getting this error message UnetLoaderGGUF This model is not currently supported - (Model architecture not allowed for conversion! (i.e. reference VS diffusers format)) But I did a clean install.... using 1 click bat

Bob Doyle - VR Films

Trying only the inpainting, I just keep getting TextEncodeQwenImageEditPlus mat1 and mat2 shapes cannot be multiplied (784x1280 and 3840x1280). I have Resize image turned off, and all models are set just like yours.

Julien Natter

I was confused not seeing the comfyUI folder after the installation, turn out the folder was in sytem32 folder for some reason

Aitrepreneur

F 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. Just go inside the python embedded folder, click on the folder path, type cmd press enter, this will bring the command prompt window and there type: python.exe -m pip uninstall torch python.exe -s -m pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128 --->replace cu128 by your cuda version (if you don't know your cuda version just open a new cmd window and type nvcc --version it will then tell you at the end Cuda compilation tools, release 12.8, V12.8.93 or something different like 12.6, in that case just use cu126 instead of cu128)

Aitrepreneur

what's your GPU?

Aitrepreneur

portable version should work right out of the box for the 5000 series as well though :/

Cristi Tanase

Oh, apparently the problem can be fixed if you change the set "COMFY_VER=v0.3.6" to set "COMFY_VER=v0.3.60" in the .bat installer file.

Erwan Cozic

did you know why i have this error? : ______________ VAEEncode CUDA error: no kernel image is available for execution on the device CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions. --------------------- i have a 5070ti

kaudlol

Yeah getting this same thing. Tried turning it off and running, which works, but the image output is very bad and doesn't represent much of the original.

Darkwan75

Well I almost got it working. Unfortunately the controlnet node can't be turned off because any position you put the switch complains that controlnet is off, yet somehow also on.

Aitrepreneur

yeah that can happen yes, for some custom nodes it can break and conflict with other nodes, hence why I put that it's preferable to start a new install check if everything works

Aitrepreneur

I got that issue as well but it's actually a browser issue, you probably got some extensions that are making that button disapear (like adblock or something like that) if you open the url with another browser in a private navigation, you'll see the button again

Aitrepreneur

I guess it could be easily adapted to a linux install but I don't have linux so I can't make sure it works beforhand

Cristi Tanase

I have used QWEN-IMAGE-EDIT-PLUS-COMFYUI-MANAGER_AUTO_INSTALL and after I'm getting missing custom nodes error. When I try to update the missing nodes I'm getting this error: Alert ComfyUI is outdated, so some built-in nodes cannot be used. together with this: Failed to find the following ComfyRegistry list. The cache may be outdated, or the nodes may have been removed from ComfyRegistry. comfy-core comfyui-aspect-ratio-crop-node comfy-core comfy-core Is there a solution for this issue? I have no VPN installed.

Tommario Minor

ImportError: IMPORTANT: PLEASE READ THIS FOR ADVICE ON HOW TO SOLVE THIS ISSUE! Importing the numpy C-extensions failed. This error can happen for many reasons, often due to issues with your setup or how NumPy was installed. We have compiled some common reasons and troubleshooting tips at: https://numpy.org/devdocs/user/troubleshooting-importerror.html Please note and check the following: * The Python version is: Python3.13 from "G:\ComfyUI_windows_portable\python_embeded\python.exe" * The NumPy version is: "2.1.3" and make sure that they are the versions you expect. Please carefully study the documentation linked above for further help. Original error was: cannot import name '_get_promotion_state' from 'numpy._core._multiarray_umath' (G:\ComfyUI_windows_portable\python_embeded\Lib\site-packages\numpy\_core\_multiarray_umath.cp313-win_amd64.pyd) Press any key to continue . . .

Manic

I am getting this Error when I try to run it with my GPU C:\AI\qwen\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\cuda\__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. How do I fix this?

Tom Schniedel

Just set up ComfyUI on an RTX 5090... had to manually install the right PyTorch (CUDA 12.1) to get it running. Would be super nice if future releases or docs mentioned a prebuilt option for 5090 cards. :P thanks for the work btw

Tommario Minor

I understand, and that's exactly what i Did. So now Qwen image edit workflow is works fine but nothing else does! So now i cant use vibevoice. So i rerun the one click installer for vibevoice and then the qwen image workflow that was just working fine is now missing nodes

Troy Walton

RE Runpod: I notice the comfy manager seems to get uninstalled ( or at least never show up in the GUI again ) after following the first "update comfyui and restart it" instructions step, making any step during and after the auto installer not really do anything but inform me I'm missing a dozen custom nodes in the workflow :/ Goes without saying I've also never seen any option to select Nightly build, since I assume that's not native to the default manager-less comfyui. Any ideas what I'm doing wrong? I'm using the aitrepreneur/comfyui:2.3.5 template with the RTX 4090. I actually had the same problem with the older Qwen workflow, but weeks ago I could at least get it working by just following the instructions.

NIghtmareQueenJune

Is it possible to use the runpod installer for installing it locally on my Linux machine? Since .bat files are windows specific I always have to install your workflows manually. If I'd be able to use the runpod installers it would make installing way easier.

Mahmoud Alhashim

this is great. Can you add the best image generator so it will be complete

Aitrepreneur

depends on what error exactly but that can happen yes, for some custom nodes it can break and conflict with other nodes, hence why I put that it's preferable to start a new install check if everything works

Tommario Minor

Am I the only one whose comfy setup breaks every time I use a new one click installer? I mean now of the previous workflows which all worked fine last night now work. suddenly im missing all sorts of nodes that were previously there. I mean i was running the tts workflow fine last night. attempt to use the Qwen 1 click installer and now nothing works not even comfy. just wondering if im alone with this experience.

Aitrepreneur

have you used the installer and updated comfyui like I showed?

Sonic

Pues me da error cuando entro: Missing Node Types y una lista enorme.

Aitrepreneur

This is a internet connection issue or vpn/proxy/antivirus blocking the download. Try disabling any of those above and try again, if need be you can always dm me