Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

WAN ANIMATE 2.2 V2 IMAGE TO VIDEO TO VIDEO KING! (Patreon)

Published:
2025-10-18 16:20:41
Edited:
2026-09-03 01:05:25
Imported:
Tags:

Downloads

Content

EDIT 3 September:NEW RUNPOD CUDA 13 TEMPLATE!

I’ve spend the last few days, completely rebuilding my RunPod ComfyUI template from scratch.... The previous template was getting very outdated and was starting to cause problems with newer GPUs, ComfyUI versions and recent workflows, so please use this new template from now on.

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

IMPORTANT: When choosing your GPU, use the CUDA 13.x filter on RunPod

The old RunPod template is now completely retired, so use this one for all new installs going forward.To use the installer use this command:bash "WAN2_2-ULTRA-AUTO_INSTALL-RUNPOD-NEW.sh"I WILL SPEND THE NEXT FEW DAYS UPDATING ALL PREVIOUS RUNPOD INSTALLERS TO THE NEW TEMPLATE SO IF THERE IS AN INSTALLER THAT I FORGOT LET ME KNOW.EDIT: COMMON ERROR AND FIXE PLEASE READ:"self and mat2 must have the same dtype, but got Half and Float8_e4m3fn"it is a wanvideowrapper node update issue that is not up to date

go into the ComfyUI_windows_portable\ComfyUI\custom_nodes folder

and delete the ComfyUI-WanVideoWrapper folder

then if you click on the folder path, type cmd press enter, it will open a cmd window in that folder and inside you can type:

git clone https://github.com/kijai/ComfyUI-WanVideoWrapper

..\..\python_embeded\python.exe -m pip install -r ComfyUI-WanVideoWrapper\requirements.txt

it will clone the latest version of the node and install the requirements, then you can reopen comfyui and redrag and drop the workflow in again.PoseAndFaceDetection

[ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running QuickGelu node. Name:'/model.0/act/Mul/QuickGeluFusion/' Status Message: CUDA error cudaErrorNoKernelImageForDevice:no kernel image is available for execution on the deviceIt's usually a 5000 series GPU issue (thanks Nvidia), so for those who have the same error, you can try this:

go inside the ComfyUI_windows_portable\python_embeded folder, click on the folder path, type cmd press enter, this will open a cmd window inside that folder and there type:

python.exe -m pip uninstall -y onnxruntime-gpu

python.exe -m pip install --upgrade "onnxruntime==1.20.1"

finish the install and then relaunch comfyui if everything else was installedHey everyone! I've created a 1-click installer for "THE NEW OPEN-SOURCE IMAGE TO VIDEO TO VIDEO KING", called "WAN ANIMATE 2.2 V2". And it’s AN AMAZING VIDEO MODEL that lets you replace ANY character in a video with just a single reference image… or take the motion from one clip and transfer it perfectly onto another character WITH INSANE PRECISION AND with my new workflow, you'll be able to replace ANYTHING inside a video, character, outfits, objects, animals, etc...the sky is the limit!

You can check out the video right here: https://youtu.be/apd68jTrxYc

The workflow: https://www.patreon.com/posts/141515195

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

LOCAL INSTALL!!! RECOMMENDED TO START FROM SCRATCH SINCE YOU NEED THE LATEST COMFYUI TORCH VERSION + SAGE ATTENTION AND TRITON INSTALLED FOR THE BEST SPEED POSSIBLE!

Download the WAN-ANIMATE-COMFYUI-MANAGER_AUTO_INSTALL.bat

Run the bat file

Wait for the install to finish

The webui will launch, load the WAN_ANIMATE_ULTRA_WORKFLOW.json file and there you go😎

DON'T FORGET TO GO UPDATE COMFYUI BY GOING INSIDE THE UPDATE FOLDER AND RUNNING THE COMFYUI UPDATE BAT FILE!!!!

IF YOU ALREADY HAVE COMFYUI INSTALLED USING A PREVIOUS INSTALLER:

Download the WAN-ANIMATE-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.

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)IF YOU ALREADY HAVE COMFYUI INSTALLED AND JUST NEED THE SAGEATTENTION AND TRITON TO INSTALL:Download the install_triton_and_sageattention_auto.bat and place it in the ComfyUI_windows_portable\python_embeded folder

Run the bat file

Wait for the install to finish

???

Profit 😎

IF YOU ARE USING RUNPOD:

Create an account if you haven't already: Runpod

When 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 WAN-ANIMATE-AUTO_INSTALL-RUNPOD-NEW.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 "WAN-ANIMATE-AUTO_INSTALL-RUNPOD-NEW.sh"

Wait for everything to be installed

Then go to the port 8188

Load the WAN_ANIMATE_ULTRA_WORKFLOW.json workflow and there you go😎

Also use the command:

tail -f comfyui.log

inside the workspace/logs to get the real time logs

PS: Once again, sorry for being quiet the last few weeks, you guys know the drill by now, if not: A Quick Update on Health, Support & Gratitude ❤️ | Patreon

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

‍​​

install_triton_and_sageattention_auto.bat (virus detected...)

Aitrepreneur

False positive it's a bat file you can open it and see what's inside

‍​​

Chrome block download ( I can't download)

Emmanuel Bronson

Can I do a second install of comfyui and would you recommend that? I'm wondering if i leave my current one just in case and install this in another drive

Aitrepreneur

click the small arrow next to “Discard” and select “Keep”. Then it’ll ask for confirmation: choose “Keep anyway”.

Mr. Friendly

I do that all the time with no issue, I even do the same drive, just another folder.

Aitrepreneur

Yes you can have multiple installs, I got like 7 myself lol

Mr. Friendly

Is this part supposed to be open for everyone? Shoudn't this just be for paying members? Also, is there a way to make it work for just images? Love the pose transfer and mask for images.

newtoai

Hello. I am having this error message when using install_triton_and_sageattention_auto.bat to install them === Detecting Torch and CUDA === 'C:\Users\marco\Downloads\ComfyUI_windows_portable\python_embeded\python.exe" -c "import' is not recognized as an internal or external command, operable program or batch file. [ERROR] Could not import torch. Please launch ComfyUI once or install torch in this python. Press any key to continue . . .

Salihz

I am having this error with the auto installer PoseAndFaceDetection [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running QuickGelu node. Name:'/model.0/act/Mul/QuickGeluFusion/' Status Message: CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed Any idea where this could come from ?

Aitrepreneur

it is for paying members, for the images you could technically use it by maybe inputting a frame cap of 1 so that it only takes the first frame of the video to transfer it to the image

Emmanuel Bronson

Grok helped me fix I had to drop down a version to 1.20 I think.

Skratata

Same issue, 5080 GPU

Mr. Friendly

Okay thanks, but just so you know I have been, but am not currently paying (will be again in the future too, but not at the moment), but I can see and download everything here. So something is odd. Just so something isn't off in your settings. Looked at all your older too, they are still open for me, even though I have become a free member (at least patreon says that I am). Tried others I been a member too and there I am blocked.

HealingPaint

Using SDPA first attempt, just based two things loaded on top (video clip on left, image on right), and get this "self and mat2 must have the same dtype, but got Half and Float8_e4m3fn" -- ideas?

XJollyboyx

unlike qwen image edit, here i just dropped the workflow in to comfy and everything works out of the box, nice. i have no idea what im doing right now

_i_

I had a lot of issues with the install_triton_and_sageattention_auto.bat script. All the issues were related to the syntax in the for loop on line20. It got too complex for me to figure out but I knew what the script was needing so I hacked it to get it to work. If you are getting errors saying you don't have Torch installed but you do, here is how i fixed it: replace the entire for loop on line 20 of the script with this: for /f "tokens=*" %%A in ('"%PY%" detect_torch_cuda.py') do ( then create a file in the python_embeded folder named detect_torch_cuda.py inside that detect_torch_cuda.py file include this bit of python import torch print(torch.__version__) print(getattr(torch.version, "cuda", None) or "unknown") save the file and run the original install_triton_and_sageattention_auto.bat as normal - it should correctly pick up on your existing Torch installation - after that I had no issues. Hope this helps someone else :)

_i_

Alternatively if you are getting errors saying you don't have Torch installed but you do, here is how i fixed it: replace the entire for loop on line 20 of the original script with this: for /f "tokens=*" %%A in ('"%PY%" detect_torch_cuda.py') do ( then create a file in the python_embeded folder named detect_torch_cuda.py inside that detect_torch_cuda.py file include this bit of python import torch print(torch.__version__) print(getattr(torch.version, "cuda", None) or "unknown") save the file and run the original install_triton_and_sageattention_auto.bat as normal - it should correctly pick up on your existing Torch installation - after that I had no issues.

newtoai

Thanks for willing to help me :-) Please see below my feedback. I Saved your uploaded file into the ComfyUI_windows_portable\python_embeded folder and then I run install_triton_and_sageattention_auto.bat and I got this below: === Detecting Torch and CUDA === Torch version: 2.9.0+cpu CUDA version : unknown [WARN] CUDA not detected from runtime. Will default later. CUDA short tag: === Upgrading pip === Requirement already satisfied: pip in c:\users\marco\downloads\comfyui_windows_portable\python_embeded\lib\site-packages (24.3.1) Collecting pip Downloading pip-25.2-py3-none-any.whl.metadata (4.7 kB) Downloading pip-25.2-py3-none-any.whl (1.8 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.8/1.8 MB 19.1 MB/s eta 0:00:00 Installing collected packages: pip Attempting uninstall: pip Found existing installation: pip 24.3.1 Uninstalling pip-24.3.1: Successfully uninstalled pip-24.3.1 Successfully installed pip-25.2 === Installing Triton === Collecting triton-windows<3.6 Downloading triton_windows-3.5.0.post21-cp312-cp312-win_amd64.whl.metadata (1.8 kB) Downloading triton_windows-3.5.0.post21-cp312-cp312-win_amd64.whl (47.3 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47.3/47.3 MB 79.2 MB/s 0:00:00 Installing collected packages: triton-windows Successfully installed triton-windows-3.5.0.post21 === Picking SageAttention wheel === [WARN] CUDA not matched. Defaulting to cu128 post4 for Torch 2.9+ Installing SageAttention: https://github.com/woct0rdho/SageAttention/releases/download/v2.2.0-windows.post4/sageattention-2.2.0+cu128torch2.9.0andhigher.post4-cp39-abi3-win_amd64.whl Collecting sageattention==2.2.0+cu128torch2.9.0andhigher.post4 Downloading https://github.com/woct0rdho/SageAttention/releases/download/v2.2.0-windows.post4/sageattention-2.2.0+cu128torch2.9.0andhigher.post4-cp39-abi3-win_amd64.whl (12.3 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 12.3/12.3 MB 64.1 MB/s 0:00:00 Installing collected packages: sageattention Successfully installed sageattention-2.2.0+cu128torch2.9.0andhigher.post4 === Optional Python 3.13 include/libs for Triton === Downloading python include/libs... % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 0 100 542k 100 542k 0 0 1537k 0 --:--:-- --:--:-- --:--:-- 1537k === Verifying imports === [ERROR] Post-install import failed. Share the error lines above. Also please see below. I own a 5090 nvidia card. When I install the ComfyUI through any of your installers, I then run the update_comfyui.bat from your created update folder, I then receive these error messages: 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. mediapipe 0.10.14 requires protobuf<5,>=4.25.3, but you have protobuf 6.33.0 which is incompatible. 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. mediapipe 0.10.14 requires protobuf<5,>=4.25.3, but you have protobuf 6.33.0 which is incompatible. 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. mediapipe 0.10.14 requires protobuf<5,>=4.25.3, but you have protobuf 6.33.0 which is incompatible. opencv-contrib-python 4.12.0.88 requires numpy<2.3.0,>=2; python_version >= "3.9", but you have numpy 2.3.4 which is incompatible. opencv-python 4.12.0.88 requires numpy<2.3.0,>=2; python_version >= "3.9", but you have numpy 2.3.4 which is incompatible. opencv-python-headless 4.12.0.88 requires numpy<2.3.0,>=2; python_version >= "3.9", but you have numpy 2.3.4 which is incompatible. torchaudio 2.5.1+cu124 requires torch==2.5.1+cu124, but you have torch 2.9.0 which is incompatible. torchvision 0.20.1+cu124 requires torch==2.5.1+cu124, but you have torch 2.9.0 which is incompatible. I truly appreciate your help on top of this great work you did and keep doing with sharing your work and knowledge :-)

Phillip

same problem .. hmm

Aitrepreneur

it's because there is a precise expiration date, since you paid in september you still have access until the same date in october ;)

Aitrepreneur

I'm working on it, it's once again a 5000 series gpu issue....... freaking nvidia...

Aitrepreneur

please dm me so that I can follow up the convo

Aitrepreneur

for those who have the same error, for now you can try this: go inside the ComfyUI_windows_portable\python_embeded folder, click on the folder path, type cmd press enter, this will open a cmd window inside that folder and there type: python.exe -m pip uninstall -y onnxruntime-gpu python.exe -m pip install --upgrade "onnxruntime==1.20.1" finish the install and then relaunch comfyui if everything else was installed

Aitrepreneur

I've already updated the installer, people shouldn't have this issue anymore

Aitrepreneur

Well perfect then, make sure you got all of the models installed though

Nicolas

After installing node dependencies, adding the models and running default workflow settings I had: WanVideoSampler self and mat2 must have the same dtype, but got Half and Float8_e4m3fn Wan model is Fp8 and Lora BF16. Are the models on the workflow ok ?

theworldofanatnom

im getting that same error if you guys ever figure it out

Aitrepreneur

can you guys send screenshots of the parameters you're using?

Aitrepreneur

can you send screenshots of the parameters you're using?

R4

-------- Checking prerequisites -------- 7-Zip not found. Downloading... % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 162 100 162 0 0 271 0 --:--:-- --:--:-- --:--:-- 271 The system cannot execute the specified program. 7-Zip install failed. Install it manually then rerun this script. Press any key to continue . . .

Phillip

and it works thank you

ti

PoseAndFaceDetection [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running QuickGelu node. Name:'/model.0/act/Mul/QuickGeluFusion/' Status Message: CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed Any idea?

So oN

Runpod seemingly has issues atm saying got no cuda device no matter if 4090 rtx 6000 or 5090 websearch for troubleshooting yielded a couple of posts of people running into the same issue it seems hope they solve it quickly can't wait to test this new workflow

8696969

Same issue. 5090.

8696969

That's a pretty big downgrade from 13.0.1. 2025 to 2022

Malachi Dolan

Hi I have a 4070 TI with 16 gigs of vram, I did a fresh comfy UI install using the automatic installer and I didn't get any errors, but when I try to do a run at 480×832×81f I run out of memory right away even with block skip set to 40, dumping my log in chat GPT it says that I don't actually have Triton or Sage attention running and that it's falling back to PyTorch attention, my question is should I be able to run this with 16 gigs of vram and the problem is that I don't have Sage attention running or is that just impossible and I need to get the quantized version of the model? editi just tried 16 frames at 240x416 and it still ran out of mem

Virtamouse

Same here. Installed from scratch

8696969

I downgraded my CUDA to 12.0 and it solved the issue. Look up NVIDIA CUDA Toolkit 12.0

8696969

Aitrepreneur also replies to another of these issues with this solution: "for those who have the same error, for now you can try this: go inside the ComfyUI_windows_portable\python_embeded folder, click on the folder path, type cmd press enter, this will open a cmd window inside that folder and there type: python.exe -m pip uninstall -y onnxruntime-gpu python.exe -m pip install --upgrade "onnxruntime==1.20.1" finish the install and then relaunch comfyui if everything else was installed"

Aitrepreneur

have you tried updating the nodes? Click on the manager button then click update all, restart and then redrag and drop the workflow in again

Aitrepreneur

have you tried the installer to install sage attention and triton first? also make sure that you have the latest nvidia drivers and that you have sysmem fallback enabled in the nvidia parameters. also how much ram you got (not VRAM, ram)? cause you usually need a huge amount of Ram to compensate for the lack of vram

Dileafino Experience

Can you help me ? I'm trying to install on runpod, but, for some reason, the manager button doesn't show after the update and restart, i'm using the 4090

R4

Unsupported 16-Bit Application The program or feature "\?\C:\Users\Gopal\Downloads\git-setup.exe" cannot start or run due to incompatibility with 64-bit versions of Windows. Please contact the software vendor to ask if a 64-bit Windows compatible version is available.

Aitrepreneur

Yeah it happens, it's apparently a browser extension issue, open in another browser or in incognito mode with the same url and it will work

R4

(after installing 7zip and then running 1click installer)

Roky86

you also need to add the "C:\Program Files\7-Zip" to your system variable "PATH" > Edit > C:\Program Files\7-Zip\ > Save

R4

I just did and its still popping up that way. am i supposed to prepare it?

R4

Add it to what?

Dileafino Experience

It worked ! Thx, I was using the edge (For some reason, on google didn't work too, but I opened into incognito mode and worked ^^)

Malachi Dolan

I ran the Sage attention installer but it said I already had Triton and Sage attention installed it did install a wheel but running my last log through chat GPT says Sage attention isn't active and that it's falling back to PyTorch attention like before, I only have 32 GB of RAM could that be the problem?

GlobalGrooveWORLD

I can't find where you have the installer for TRITON?

Emmanuel Bronson

is it loading in the start of comfyui? Look at the logs. If not need to add it to the launcher in text editor

troy X

Does anyone here have an issue with the wan video sampler? it runs, then faults out with this.Command '['C:\\ComfyUI\\Wan 2.2 animate\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\runtime\\tcc\\tcc.exe', 'C:\\Users\\JORGEN~1\\AppData\\Local\\Temp\\tmp96s1hq1k\\cuda_utils.c', '-O3', '-shared', '-Wno-psabi', '-o', 'C:\\Users\\JORGEN~1\\AppData\\Local\\Temp\\tmp96s1hq1k\\cuda_utils.cp313-win_amd64.pyd', '-fPIC', '-D_Py_USE_GCC_BUILTIN_ATOMICS', '-lcuda', '-lpython313', '-LC:\\ComfyUI\\Wan 2.2 animate\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib', '-LC:\\ComfyUI\\Wan 2.2 animate\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib\\x64', '-IC:\\ComfyUI\\Wan 2.2 animate\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IC:\\ComfyUI\\Wan 2.2 animate\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IC:\\Users\\JORGEN~1\\AppData\\Local\\Temp\\tmp96s1hq1k', '-IC:\\ComfyUI\\Wan 2.2 animate\\ComfyUI_windows_portable\\python_embeded\\Include']' returned non-zero exit status 1.

Travis Hightower

Hi there. Trying again to get this installed for the first time.. I am still getting this error when I try to run the installer. I apologize if it has already been addressed elsewhere

Travis Hightower

Is there a pre-required step I am missing before I try to run the WAN-ANIMATE-COMFYUI-MANAGER_AUTO_INSTALL.bat? I think I saw someone else with this problem last time, but not sure if there was an answer. I get this error pop up. Unsupported 16-Bit Application × The program or feature "\ ?? \C:\Users\Travis and Jennifer\Downloads\git-setup.exe" cannot start or run due to incompatibity with 64-bit versions of Windows. Please contact the software vendor to ask if a 64-bit Windows compatible version is available.

troy X

You must have an older GPU. This is a similar error on my 2080Ti

Larry1710

Not seen this one before - running Wan Animate 2.2 Auto_Install. Errors on git-setup.exe - "cannot start or run due to incompatibility with 64 bit version of windows". I've run other Installers b4 with no problems on another system that was WIN 11 Pro 64 - what did i break? 🙁

kaudlol

[[ FIX FOR THIS ERROR ]] Obviously, update all necessary nodes, etc., but also the WanVideo Wrapper needs to be on nightly. Switched to nightly and finally got an output. Kijai's comment response to error: "I know, and I fixed it soon after in the code, I had the same error when I tried, so your nodes still aren't up to date, make sure they are on nightly version currently."

Alpha Ghost 47

For some reason once the ui loads up i can move around, the small key at bottom right is telling me what I am doing but the actualy comfy ui isnt moving??? (So I can't navigate or zoom in)

E

flow worked for me but my image generation is super grainy. Any suggestions?

Peter L

I'm also getting this error. 5090 as well.

HealingPaint

I'm just using default workflow parameters. I noticed upon opening the workflow though that is was set to MASK-SEC, and not MASK. I changed that back after if failed the first time and I noticed, that, but it didn't help. settings1: https://drive.google.com/file/d/1YspvVqNf1HoHquZ61rSwL1r0PH0sWorl/view?usp=drive_link settings2: https://drive.google.com/file/d/1ZmXqDaBGU8Q6Sp9tLxDts8aVDbjaFwES/view?usp=drive_link error: https://drive.google.com/file/d/1MBQTIhO7SJ69MJnFpN7GBSvYjXd7l3gW/view?usp=drive_link I am on nightly version BTW

troy X

Nevermind. everything worked after turning off triton.

Emilien-Edmond D'aramitz (LeonceTheLion)

In case anyone is getting this error in runpod when trying to install the .sh file: "git: 'lfs' is not a git command. See 'git --help'. The most similar command is log" here is a solution that Gemini helped me fix: Even though you are on a specialized RunPod template (like the ComfyUI one), sometimes these base images do not include Git Large File Storage (LFS) by default, especially if they are trying to keep the initial image size small. The script you are running, WAN-ANIMATE-AUTO_INSTALL-RUNPOD.sh, is clearly trying to clone a repository that uses Git LFS to manage the large model files, and it's failing because the lfs command is missing. Since you are running as root in a Linux environment (likely Debian/Ubuntu-based), the fix is simple: install Git LFS. Here are the commands you should run in your RunPod terminal (JupyterLab terminal or web terminal): Update your package lists: Bash apt update Install Git LFS: Bash apt install git-lfs -y Initialize Git LFS (Good practice): Bash git lfs install Re-run your installation script: Bash ./WAN-ANIMATE-AUTO_INSTALL-RUNPOD.sh After running these steps, the script should execute without the git: 'lfs' is not a git command error, and the necessary large files for WAN-ANIMATE should be downloaded successfully.

newtoai

this worked for me and I also have a 5090. Thanks for bringing this solution up!

Sense Departed

download git from git-scm.com, the one click downloader didnt help me with that one, had to do it manually

evil chucky69

you know i must have installed comfyui, both portable and std versions, on my pc, at least 12 times. i have to say not once, not ONCE - have i been able to get it operating without some kind of crap that needs updating, installing, re-installing, downloading....blah, blah, blah. missing nodes, missing this and that, outdated this, change the version to this...blah, blah, blah...what a piece of shit program that promises sooooo much, yet delivers sweet fk all! the day this program gets nuked i will be dancing in the fken street with no clothes on!

Nick Sarkisov

guys, how fix this type of error: Failed to update ComfyUI or several extensions / TypeError: Failed to fetch. When I'm updating comfy.

AiArtSenpaii (edited)

Comment edits

2026-09-08 20:00 weird it says my 64 bit cant run this because its not up to date? Is all of stable diffusions crashing this month?
2025-10-19 13:01

Boxertwin

Hi, first attempt I got this message "self and mat2 must have the same dtype, but got Half and Float8_e4m3fn"

Aitrepreneur

it is a wanvideowrapper node update issue that is not up to date go into the ComfyUI_windows_portable\ComfyUI\custom_nodes folder and delete the ComfyUI-WanVideoWrapper folder then if you click on the folder path, type cmd press enter, it will open a cmd window in that folder and inside you can type: git clone https://github.com/kijai/ComfyUI-WanVideoWrapper ..\..\python_embeded\python.exe -m pip install -r ComfyUI-WanVideoWrapper\requirements.txt it will clone the latest version of the node and install the requirements, then you can reopen comfyui and redrag and drop the workflow in again.

Aitrepreneur

yes it is a wanvideowrapper node update issue that is not up to date go into the ComfyUI_windows_portable\ComfyUI\custom_nodes folder and delete the ComfyUI-WanVideoWrapper folder then if you click on the folder path, type cmd press enter, it will open a cmd window in that folder and inside you can type: git clone https://github.com/kijai/ComfyUI-WanVideoWrapper ..\..\python_embeded\python.exe -m pip install -r ComfyUI-WanVideoWrapper\requirements.txt it will clone the latest version of the node and install the requirements, then you can reopen comfyui and redrag and drop the workflow in again.

Aitrepreneur

it's the install_triton_and_sageattention_auto.bat file

Aitrepreneur

I do recommend installing triton though, use the install_triton_and_sageattention_auto.bat installer

Aitrepreneur

try uninstalling and reinstall git for windows

Aitrepreneur

try restarting comfyui and open in another browser

Aitrepreneur

Image gen? you mean video gen, what parameters are you using?

Aitrepreneur

never had this issue during testing, are you sure that you are using my comfyui template?

Aitrepreneur

Bro I cannot agree more unfortunately, believe me I wish there was an easier way, I spend so much of my own time troubleshooting shit on my side that I also get annoyed with it but unfortunately comfyui is the best more versatile open source tool there is right now, so yeah this all come to a cost....if you got any issues you can always dm me, we'll try to find a solution

Aitrepreneur

first use the update comfyui bat file located in the update folder to update comfyui, never use the manager

Aitrepreneur

it is a wanvideowrapper node update issue that is not up to date go into the ComfyUI_windows_portable\ComfyUI\custom_nodes folder and delete the ComfyUI-WanVideoWrapper folder then if you click on the folder path, type cmd press enter, it will open a cmd window in that folder and inside you can type: git clone https://github.com/kijai/ComfyUI-WanVideoWrapper ..\..\python_embeded\python.exe -m pip install -r ComfyUI-WanVideoWrapper\requirements.txt it will clone the latest version of the node and install the requirements, then you can reopen comfyui and redrag and drop the workflow in again.

newtoai

C:\Users\marco\Downloads\ComfyUI_windows_portable\ComfyUI\custom_nodes>git clone https://github.com/kijai/ComfyUI-WanVideoWrapper C:\Users\marco\Downloads\ComfyUI_windows_portable\python_embeded\python.exe -m pip install -r ComfyUI-WanVideoWrapper\requirements.txt error: unknown switch `m' usage: git clone [] [--] [] what I doing wrong?

Larry1710

Very Possible new build this week on Win 11 Pro 24h2 - 25h2 is already "pending" can't wait :{ Anybody else seeing this on recent builds? As MS will force the updates it's not going away - hope there's a work-around if it is an MS issue. LOVE the AUTO INSTALLERS!

Travis Hightower

thank you - so now I need to figure out how to do it manually lol - I was able to download git. Its basically a command window? EDIT: Oh with get installed it just works.... I think . .. interesting. Thank you!

Travis Hightower

ok - so everything seems to work except when I try to run a template, a bunch of models are missing. I can download them, but don't know where to put the files once I do.

EastCoast Logos and Designs

this is amazing! But having an issue with some nodes. I tried updating and I cant seem to get them: PoseAndFaceDetection DrawViTPose OnnxDetectionModelLoader

Phillip Gimmi

the workflow you reference does not work in runpod

E

you are right video gen. WAN 2.2 I2V Parameter Dump UNet Loader (GGUF) unet_name: Wan2.2-I2V-A14B-HighNoise-Q5_K_S.gguf UNet Loader (GGUF) unet_name: Wan2.2-I2V-A14B-HighNoise-Q5_K_S.gguf CLIP Loader (GGUF) clip_name: umt5-xxl-encoder-Q5_K_S.gguf type: wan Load VAE vae_name: wan_2.1_vae.safetensors 🎬 WAN Image to Video width: 512 height: 512 length: 81 batch_size: 1 🎛️ KSampler (1st Stage – High Noise) add_noise: enable noise_seed: 874220123881467 control after gen: randomize steps: 20 cfg: 1.5 sampler_name: euler scheduler: simple start_at_step: 0 end_at_step: 2 return_with_leftover latent: enable 🎚️ KSampler (2nd Stage – Refinement) add_noise: disable noise_seed: 828934133109108 control after gen: randomize steps: 20 cfg: 10.0 sampler_name: euler scheduler: simple start_at_step: 2 end_at_step: 4 return_with_leftover latent: disable 🎞️ Video Combine (VideoHelperSuite) frame_rate: 8 loop_count: 0 filename_prefix: AnimateDiff format: image/gif pingpong: false save_output: true

Sequoia

Thanks a lot! is it normal for 1 second of Video to take 1.5 minutes to be generated (25 frames)? (RTX3090)

Vincent Nestler

Here is my current issue - which may be that I am on a 5090 laptop: PoseAndFaceDetection [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running QuickGelu node. Name:'/model.0/act/Mul/QuickGeluFusion/' Status Message: CUDA error cudaErrorNoKernelImageForDevice:no kernel image is available for execution on the device Thoughts?

N

Hi, I used the provided Triton/SageAttention installer, but I get this error when I have them enabled. I'm running on a 4090-24vram with 96ram on an Intel i9. Before I ran the Triton/Sage installer, I was getting the same Python error, except after I ran it, the error about TorchDynamo started happening as well. If I disable Triton and use SDPA it'll run. Any suggestions? Thanks! CalledProcessError: Command '['H:\\Comfyui_WAN_Video\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\runtime\\tcc\\tcc.exe', 'C:\\Users\\conta\\AppData\\Local\\Temp\\tmp_ufg_on6\\cuda_utils.c', '-O3', '-shared', '-Wno-psabi', '-o', 'C:\\Users\\conta\\AppData\\Local\\Temp\\tmp_ufg_on6\\cuda_utils.cp313-win_amd64.pyd', '-fPIC', '-D_Py_USE_GCC_BUILTIN_ATOMICS', '-lcuda', '-lpython313', '-LH:\\Comfyui_WAN_Video\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib', '-LH:\\Comfyui_WAN_Video\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib\\x64', '-IH:\\Comfyui_WAN_Video\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IH:\\Comfyui_WAN_Video\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IC:\\Users\\conta\\AppData\\Local\\Temp\\tmp_ufg_on6', '-IH:\\Comfyui_WAN_Video\\ComfyUI_windows_portable\\python_embeded\\Include']' returned non-zero exit status 1. Set TORCHDYNAMO_VERBOSE=1 for the internal stack trace (please do this especially if you're reporting a bug to PyTorch). For even more developer context, set TORCH_LOGS="+dynamo"

Kōmei Sanga Artist

check the description here again, he has notes on that error, just below the first error listed

James

Any workaround for runpod yet? Im having the same issue but there aren't any available GPU's with the Cuda version to get it working.

Ricky Lawrence

Did you know that in the video, the mask is connected to...? but in the workflow json file, it is not? where does it connect to?

Ricky Lawrence

Did you know, everybody environment is different. Clean install ComfyUi with the person whom you want the same results, otherwise, it's a hit and miss.

H-BOX

Hi, it's amazing that you made this. but on the runpod, even after upgrading everything, it doesn't solve this time. Here are the missing nodes that doesn't fix : OnnxDetectionModelLoader PoseAndFaceDetection DrawViTPose

newtoai

were you able to avoid having grainy videos? Also they are chopped sometimes at the top and bottom

Larry1710

BOOM! that was it - resolved some git issues earlier today, good to go!

Tom Baron

PoseAndFaceDetection cannot convert float NaN to integer. Anyone else getting this error? UPDATE GPT told me it was my video. "on frame 296 of your input video, no face was detected. When it tries to compute a bounding box from missing face data, some math divides by zero, which creates NaN values—and that's what crashes the whole workflow."

E

the loader didnt dl specific models for me. Specfically low noise and high noise models. Once i added that and added the light lora things improved a lot. Could still be better quality though

Khajiirah
>>190932634

If you cant find VitPose Custom Node, its called "ComfyUI-WanAnimatePreprocess" from Kijay, in ComfyUI Manager to download and install. After installing this you get the three Nodes. After i downloaded the "ComfyUI-WanAnimatePreprocess" with the Manager and restarted ComfyUI, it worked and all three missing Nodes where available and nomore Red. =^..^= Have Fun! If it won't work afterwards, (does nothing when clicking Start), make sure to enable the Mask Group (down below), as inside the Masking Group is the Export File inside (combine and save video), without it, it does nothing as it does not know where to export and save the file.

Khajiirah
>>190943170

Sorry i did missunderstand, so i deleted my comment. I wrote something like "try to replace the buggy node wich gives you the error, not copy it!" But your Problem is different, i will look into that. It could mean, that inside your video, the characters head turns away and so it can not recognize the face correctly.

Aitrepreneur

are we talking about the same workflow here?

Aitrepreneur

I can't replicate the issue, I just did a new install on runpod and everything works as intended

Aitrepreneur

I can't replicate the issue, I just did a new install on runpod and everything works as intended maybe it was a temporary runpod issue

Aitrepreneur

local or runpod? if it's local just use the WAN-ANIMATE-MODELS-NODES_INSTALL.bat file inside the comfyui folder or just install the ComfyUI-WanAnimatePreprocess custom node For runpod, try again, it might have been a temporary runpod issue because everything works now with the installer

EastCoast Logos and Designs

Local and I will try that! And just a question outside this workflow do you have multiple versions of comfy. I have like 4 already haha

CodeCrafters Collective

same error and same gpu and cpu, used the install_triton_and_sageattention_auto.bat and installed fine.

H-BOX

Thank you for replying guys, I'll look for that node !

Einziger

This is awsome! It would be nice to have a installer with everything (Wan/Qwen/Flux...) on a regular basis. Can you do that please?

H-BOX

Hi! I somewhat made it work on vastai, reinstalling manually the missing nodes about ComfyUI-WanAnimatePreprocess, thanks for your work ! One more question : How can I change the fps ? I do not want to use the original one, because that is too much. Wan goes well starting from 16 so I changed the force fps to 16 but I accelerates the movements in final results. Any easy tweak to change that ? Also, just wanted to say it's incredible because I was using a wananimate WF I found randomly and it sucked compared to this one, didn't even take the face etc ! Thanks !!

Cristo

Would be interessting if we could upload more than one image. Let's say a person, front and back. Because with only one image the AI won't be able to guess what the image reference looks like from behind

dan

every time i try to run the WF the resize image node for the reference image halts the process and i get no generation. any ideas how to fix?

Angelika Thruston 💨

Everything in the workflow works great until I tried using the SeC masking. I followed all instructions in the video, but I get this error: SeCModelLoader Failed to load SeC model: No module named 'omegaconf' I tried downloading a fresh copy of the SeC model from the following link, but that didn't help: https://huggingface.co/Aitrepreneur/FLX/blob/main/SeC-4B-fp16.safetensors I tried (re)installing omegaconf with the following command, to no avail: pip install omegaconf This is the output: Traceback (most recent call last): File "C:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 498, in execute output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, hidden_inputs=hidden_inputs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 316, in get_output_data return_values = await _async_map_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, hidden_inputs=hidden_inputs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 290, in _async_map_node_over_list await process_inputs(input_dict, i) File "C:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 278, in process_inputs result = f(**inputs) File "C:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\Comfyui-SecNodes\nodes.py", line 435, in load_model raise RuntimeError(f"Failed to load SeC model: {str(e)}") RuntimeError: Failed to load SeC model: No module named 'omegaconf'

Lance Fanger

Will the node installer and Triton/Sage attention installers work for the base version of ComfyUI or does it have to be the portable version?

Väinämöinen

Each and every time I install on runpod it fails. I follow the instructions to the letter but the ComfyUI manager always disappears after updating and restarting. I've spend maybe 100$ trying to get it to work on runpod :(

Tinkering

First OVI was great. Lots of fun. Question. I have installed this as an addon to my existing Comfyui and as a new install. I have used videos created with OVI and local images and web based images. I am doing the exact same setup as your first demo section of the video (though I enter the video size 992/512 to match the input video. Note: I also updated comfyui from .bat file. Regardless of what I do I get. WanVideoModelLoader Error while deserializing header: incomplete metadata, file not fully covered.????

Tinkering

fixed. did a file check of the three installs on my machine and the Wan2_2Animate-14B-fp8_scaled_e4m3fn_KJ_v2.safetensors had three different sizes. Ya that would be an issue. Downloaded another from Huggingface and now it workss!!!

E the G

I'm having the same exact issue as well. Dang. I guess the speed boost won't work on my current setup. Don't feel like troubleshooting for hours today lol.

Mr Morphin Time

When using rtx 5090 or rtx 6000 ada on runpod, the manager button no longer shows up and I can't install custom nodes. Anyone know a fix for this?

newtoai

Hi Guys. Is there an active discord channel where we can all chat and help each other with throubkeshooting , as supportes for this channel?

nurtu

Im not getting the best likeness to my ref, what do you recommend besides training a lora? Up the steps?

Vaivij

Is there a way to change the mask color? If I am replacing anything black, a portion of it comes back as the mask 'blocks' that don't go away, especially at the edges.

Vaivij

Also, in the loras, do we put high noise wan2.2 loras or lownoise ones?

Pizza

Having problems with this. It worked at first, but now all of my videos that I load are skewed in the masking area down below. They are zoomed in a lot and I can't mask since it's so skewed. Do I need to match the h x w exactly with the video or picture?

Tinkering

To some of the people having issues. I have found a couple things. First, the installs are not always perfect. Run issues or missing nodes have been cleaned up by just a fresh install. I also had to check the models individually. Re-downloaded one that got corrupted. I did have Mask issues several times. First you need make sure you are following the steps exactly. I had that user error a couple times. However, even after correcting my errors I had masking mess ups. I have found that restarting the program and the web browser fixed it - most of the time. However, when that did not work, reload a clean copy of the workflow. Just had an issue where I could not get the video to mask properly. Background would appear and disappear. reboot did nothing. New video did nothing. I redownloaded the workflow. All was back to normal. Just my two cents.

James Inage

Used the triton/sage attention installer bat and it seemed to have installed properly or it said it did but every time I try to use sageattention in one of your workflows i get the error PY_SSize_T_Clean macro must be defined for "#" format Anyone know the issue or possible fix?

TheRangor

I'm having issues with the SEC masking too. The auto masking seems to work more consistently. With the SEC mask, more often than not, it will fail to completely mask the character and their face. Perhaps I'm just doing it wrong, I don't know.

whitebread

Hi, do i still need to put both low and high lora// or just only one of them

Frank B

Welp, in this workflow, Video seems to be a no go on a 3070 (8GB) I just get 'out of VRam' half way though generation, using the Q4KS gufs, I know its possible, Im sure there is a comfy node off off loading to System RAM, anyone?

newtoai

Is there a place is the workflow to change the text Encoder used?

chibialien

He mentions it in the video. If you use the WanVideo Block Swap node, that is a part of the workflow, you should be able to offload from VRAM to system RAM

Extreme Mojo

Hi and Thanks for all the work you do. I'm trying to use this workflow on runpod. I do the entire instilation as instructed, however it is not loading the following nodes: OnnxDetectionModelLoader, PoseAndFaceDetection, and DrawVitPose. The manager is not finding these three nodes when I try install missing nodes. Is there something I'm missing?

Nicholas Byford

Great work. Anybody know how to keep the black blocks from showing up in the final video when using the SEC masking?

Frank B

Thanks, I have to go back and check for it to try again, for now I got alternative Wan 2.2 checkpoint (rapid mega AIO) running, and its all good.

Mr Morphin Time
>>191049101

I was able to fix it. I didn't know how to restart comfyui in runpod without the manager button. But port 8000 has a start and stop button so I restarted comfyui from there and then manager showed up.

Emiliano

Ditto AI video editor (built on Wan, by the Wan team) dropped, hope to see a video on it soon. Our AI overlord always has insanely good workflows.

Chris Rose

Do you know what would be great for this? If you could create a version that is like your wan extended video workflow. If you could upload a video regardless of how long it is, and have it broken down into sections, vram offloads and then goes on to the next section.

JOHN

is there someone like this guy that makes 1 clicks that work?

M. Stx

thanks was having this issue too, now the 1click installer bat is workin

Tinkering

I found if that happens you need to move the green out past the face a bit and leave a bit of room between the green and red circles. also extending the mask as the video discusses help

Einziger

It did work for me with a 5090. First time I have nothing to tweak or modify.

SABER

omg this is why i joined the patreon. hoping it works for me too!

James Inage

Used the triton/sage attention installer bat and it seemed to have installed properly or it said it did but every time I try to use sageattention in one of your workflows i get the error PY_SSize_T_Clean macro must be defined for "#" format Anyone know the issue or possible fix?

Arkadiy Grigoryants

Thank you so much for your amazing work! This is honestly the best thing I’ve seen in the world of AI. I’ve been working in computer graphics for over 20 years - 3D, compositing, live-action integration - and what I was able to create using your workflow absolutely blew my mind 😍 It’s incredibly powerful and inspiring. Thank you for your effort and for everything you’re doing for the community 🫂

LouNokoAI

Hey would you be able to please help me set it up as well I am having some problems with it and I do not know how to fix them.

Fizzy

Hey I tried creating a video and for some reason I am running into the following issue: WanVideoTextEncodeCached Error while deserializing header: incomplete metadata, file not fully covered I have installed everything but the runpod as I'm not using that. I'm not quite sure how to fix this issue. I get to around 30% and have that issue. If anyone knows how to fix this please let me know

Taiga

close the running comfyui and go to ComfyUI_windows_portable/updata and run the update_comfyUI.bat file. if it's still a problem, try running the models nodes bat file.

James IV

I can never get any of these to work. I guess I need to learn actual AI coding because none of these are ever just 1 click install and load the workflow and go. Its just error after error and hour after hour of trying to fix them for nothing.

Ryan B

Wondering if anyone else is struggling with the runpod install for the latest Wan Animate 2.2 workflow? Ive been struggling aall day, restarted like 4 times now, following instructions to a T and I cannot get it working. When following instructions exactly, I end up with an updated ComfyUI interface, but cannot access the manager (ui button never appears, multiple reboots dont help. In one of my attempts I was able to get the manager to work after manually installing it (have not successfully recreated this since) however even with that working, I could not get the custom nodes from the workflow to install. In particular, WAN Vide Wrapper refuses to import, no matter what i do. The only other real valuable info as to whats happening that I have come across so far is that a whole lot of the following warning for a number of nodes: "[DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version." Any help would be appreciated, though my guess is that something got changed/updated and the install script needs updating, but IDK. Thanks in advance for any help. I love the channel and wish Aitrereneur all the best!

a pretty parrot

i made a loader script that runs everything after pod deploys. the one thing i always have to do separately is install librosa. you can do that using web terminal, this command should do it: source /workspace/ComfyUI/venv/bin/activate && pip install librosa EDIT: this is on runpod

a pretty parrot

What problems? I’ve ran this one about 5x so far on few different runpods. Double check you are using the aitrepeneur/comfy2.3.5 or whatever image, and make sure you edit to set the volumes sizes.

a pretty parrot

Are you using the airtrenepeur:comfy image?

Dmitry Izoskin

i can`t understend why one click installer doesn`t work, it makes nothing when download ...

Timbo

as always great work, on quick idea is that it would be great to have the download size in the discription somewear , just be really helpfull to know what sort of size files we are going to need to download

a pretty parrot

i hadnt looked at that workflow yet, i was thinking of that idea already. glad he already made one

DravenLock

Do these work with AMD cards yet?

gr33n

Hey everyone, I’m new to ComfyUI and trying to run it locally on my RTX 4080 Super (16GB VRAM, 32GB system RAM). I’m getting this error: torch.AcceleratorError: CUDA error: out of memory 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.

gr33n

is it even possible with this setup?

Spiritonfire

I'm running in the same problem and already did what he say's on the post. I guess it's on our own to figure this out. I will keep you update

Spiritonfire

So I got able to work. on the Wan Video Block Swap. Enable offload_img_emb to True and Offload_txt_emb to true as well.

Mark P

I have the same issue. What was the solution?

mark

having same issue

mark

Thanks. Guess you can't just throw any source video on there.

mark

sounds like you maybe already aware, but for anyone else like me, you can still use sdpa (i think is what it is called instead of sageattention)

skull rage

I can't run this because im getting triton related issues even though i have triton disabled in the workflow. have tried running with set TORCH_COMPILE_DISABLE=1 too, and i have cuda 12.8 Any ideas? Command '['C:\\WAN2.2\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\runtime\\tcc\\tcc.exe', 'C:\\Users\\dariu\\AppData\\Local\\Temp\\tmpn2_kyake\\cuda_utils.c', '-O3', '-shared', '-Wno-psabi', '-o', 'C:\\Users\\dariu\\AppData\\Local\\Temp\\tmpn2_kyake\\cuda_utils.cp313-win_amd64.pyd', '-fPIC', '-D_Py_USE_GCC_BUILTIN_ATOMICS', '-lcuda', '-lpython313', '-LC:\\WAN2.2\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib', '-LC:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v12.8\\lib\\x64', '-IC:\\WAN2.2\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IC:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v12.8\\include', '-IC:\\Users\\dariu\\AppData\\Local\\Temp\\tmpn2_kyake', '-IC:\\WAN2.2\\ComfyUI_windows_portable\\python_embeded\\Include']' returned non-zero exit status 1.

James Inage

I had tried that, it does seem to work, just not of course as fast as sage should, there's also a couple nodes out there that require sage, I do appreciate the advice though, more than @Aitrepreneur bothers to help anyone with, I have messaged them every couple days for a couple weeks now and they don't even bother to send a response, for the insane amount of money they're making posting open source scripts and workflows you would think that they would at least bother to respond. Found there's a lot of people who are posting free installation scripts that do the exact same thing that they're selling here, so canceled my membership

Morbid

anyway to run on ubuntu?

troy X

Having an issue with Runpod. After I update to ComfyUI to latest 3.68. The manager UI is gone and nodes are missing. Labels (rgthree) and Fast Groups Bypassers (rgthree). (PS I'm also running on my PC with ComfyUI 3.37 with no issue).

Xyzzx

Sweet. But Mocha is the new king!

troy X

Nevermind fixed it by running on a different computer. WOW 4090 is super fast vs my 3090.

ColinMacLaren

The specific version of the libs depends on the version of your embedded Python. Mine was 3.12, yours is 3.13 I presume.

Zzz Kshero

facing same issue, any help please?.

Exams Baba

What do you mean "fixed it by running on a different computer" on runpod?

Bimb-OH!

Hi! I got everything working and tested both masking methods with awesome results. I've got two questions though: - What's the best option to improve the likeness of the generated face? CFG? Steps? - At the very end, the video seems to play a few frames backwards. Why is that? Thanks!!

troy X

I ran Runpod on a different computer and keep restarting the port 3000 and finally one of the restart showed the comfyUI manager and no more missing nodes, but now I have the same issue again when I create a new pod.

troy X

On Runpod anyone know how to get the ComfyUI manager back after the update all and restart. It's gone and no longer on the top right corner.

troy X

Nevermind, it was a Chrome browser issue. Switch to Edge and the manager came back.

M B

Hi there, is there a possibility to change the workflow to GGUF Model of Wan2.2 etc. so it would work with my RTX3060 12GB VRAM. I installed everything like in the instructions and the workflow starts doing its thing, until i get a reconnecting message from comfy ui. In the past it meant for me that i ran out of space. Can someone help me please. Thanks a lot!!

LouNokoAI

Hello, I have installed the workflow I am very happy with it and I am glad I have the opportunity to use it, I just have a problem regarding the physics of the upper body of a person such as jiggling and realistic movements of the upper body and chest, how can I fix that? Thank you!

Ez

This no longer works for me. Originally it was great, churned out 1280x720 vids on my 4090 no issue. I can still churn them out on the image/text wan2.2 workflow (as well as having no issues on any of my own). Now though, It doesn't matter what I update, clean install, etc, or how tiny the vid and image are, it will always run out of memory.

jack2pot

Just downloaded it today, same issue torch.AcceleratorError: CUDA error: out of memory

MSUM

Same. I just get various insufficient memory errors no matter what I try, like this one below. I have a 3080 w/16GB VRAM and 64GB of system ram: [enfore fail at alloc_cpu.cpp:121]] data. DefaultCPUAllocator: not enough memory: you tried to allocate 22484736000 bytes.

newtoai

same issue here . I am wondering if the issue is with an update of windows or nvidia card drivers

Sfde12

I dont know if im doing something wrong but if i click on port 3000 for ComfyUI i get The port is not up yet. Im using a runpod with your template

Paperpotter

Same - just been trying to get it working on RunPod 5090 - No joy.

Hype

Are we not able to run this locally on lower VRAM GPU's? Currently running a 3080 with 10 GB of VRAM.

Hype

I'm running this on Arch so this definitely works on Ubuntu. The only thing is his one click-method of installation won't be helpful to you since .bat files only work nicely on windows.

Maemo

when i run Wan-animate -comyui- manager auto install.bat error pop up with this title :unsupported 16bit application : "cannot start o run due to compatibility with 64bit-bit version of windows

Loli Pot

it doesnt work . error appear ! is this a scam ? its say not compatible with 64bit

DarkVOID

Same here. Hopefully we get a response soon

Janne Kallio

Maybe I missed it, but first frame, last frame video would be nice.

S T

it takes 10 mins to generate videos with RTX4090 on runpod i havent had luck making it generate faster with the 5090 and with 6000pro i get some error about CUDA. what is the best gpu for faster video generating on runpod? I dont care about cost. THANKS IN ADVANCE!

Aitrepreneur

What cuda error? 6000 pro would be the best gpu for anything ai related yes

VixDreams

You can run it, but you need a GGUF model, which are quantized models for graphics cards with less VRAM.

VixDreams

How do I download the workflow?

Samuel Anttila

Spent over an hour troubleshooting, but everything is now working on the latest ComfyUI version. This is amazing, thank you!

Tito Negara

Hey, can you help me? i was getting cuda error out of memory, i run it with rtx 3090

caca drk

what settings can i tweak in this workflow to get the highest possible quality output? the faces are coming out a bit bland/soft... not showing enough emotion like the original videos im using. not enough eyebrow lifts, smiles, winks etc... how can i tweak this to make sure its copying the original video properly and not giving me a soft emotionless face on the output?

SebasMad Gau

Hey, with the problem PoseAndFaceDetection you mention in the post. I cant find the folder comfyui windows portable with Runpod , pls help

zodrix

same issue for me, i guess it's between the nvidia driver cuda compatibility with pytorch and cuda inside the comfyui

Avishek

where do the vitpose models go into? which folder?

Timeless1175

Hi Aitrepreneur, I am late to the party but I want to know if i add a WAN lora, how to do it? the loras i downloaded are to be used with High and Low noise in pair. so how do i do it in your workfow? thanks!

aradfuton

I just became a member to support your content and I must say it is frustrating to hunt down the workflow jsons. Please update your posts to include their respective workflow jsons.

Wallosh Wolfmann

If you use Wan2.2 you need both High and low noise lora. If you use Wan2.1 you only need the low noise one.

Stuart Galloway

the auto install bat wont run on windows 11 as it is flagging as a unsupported 16 bit application

Toxic

will this SAGEATTENTION AND TRITON install work for wan 2.2? Can you update the work flow with it?

floflo tutu

there is a problem on runpod, after i update comfyui, if i restart the comfyui, and i refresh the page it cannot load the page anymore, i have instead of comfyui: error: 403 forbidden.

Aitrepreneur

Just add a / at the end of the url then press enter