Home Creators Posts Import Register Favorites Logout

Downloads

Content

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 just released a new video on "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

I also created a 1-click installer for the Webui for my Patreon supporters for Local AND Runpod, that you can find right here: https://www.patreon.com/posts/141515205

Here the link to the manual install guide: https://pastebin.com/sjN3JzbC

Down below you will also find the WAN_ANIMATE_ULTRA_WORKFLOW.json for free for you to use!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 ❤️ | PatreonAs 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

GrowthTrack

Some of the models, such as vitpose_h_wholebody_data.bin are missing from your huggingface.

L Hill

Appreciate it and it's all good. Take care of yourself first, I don't mine support quality content.

Dawid Ogiegło

Do you folks have any suggestion what could the way to zoom out an input video a bit?

Max Schreck

Damn... This is the best WAN Animate workflow I've had so far.

Dawid Ogiegło

And if there is a way to use face and poses from two different videos?

Aitrepreneur

I don't understand what you mean by zoom out but to be able to use a different pose and face you need to modify the workflow a bit and add a video loader for the face part

Cristo

I can't zoom in or zoom out on the workflow and sometimes I can't upload a video reference.

ELITE AI EDUCATION

Hey there! I’m on a Mac and I’m thinking about joining your Patreon. Just wondering if you have any workflows that are Mac-friendly or that show how to use RunPod specifically on a Mac? Thanks

Aitrepreneur

I don't have a mac so I can't test anything on it beforehand however runpod is a website so you can already use it no problem

May Link

DID YOU MANAGE TO SOLVE IT? HOW DO I DOWNLOAD THE NODES THAT ARE MISSING?

Oscuros

I had this issue and fixed it by by updating my custom nodes, if you've installed one of the custom nodes used in this workflow before now that's where I'd start. For me, I had to update KJnodes and the WanVideoWrapper nodes.

Maja Ji

Thanks. It really works. Just one question, when it uses VIT poses, it randomly track the pose of random character, can we control that too like making it to track particular character pose from the video?

May Link

I got this error, is it lack of memory? : WanVideoSampler self and mat2 must have the same dtype, but got Half and Float8_e4m3fn

TomiTom1234

I did everything as explained in the video, but sadly I get this error in the middle of the process: 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 And I see the "Pose and Face detection" node is in a purple square, where the process stops. I checked the "Mask" and "Background_image" because I use the automatic masking. Any idea what could cause this issue? I also tried "Triton" enabled and disabled, same eror. BTW, my card is 5070 Ti , 16 gb vRAM The whole console process is in this screenshot: https://ibb.co/84GNtx4M

Freno Deckilin

I got exactly the same problem. Might the WanVideo Model (FP8) and the WanVideo Lora/WanVideo VAE ( both FP16) cause this problem?

Torsten

I am running a 5090. The workflow worked on time but now I am gettin this error message: 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.

Aitrepreneur

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

you are running out of vram for some reason, just restart comfyui and try again

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.

TomiTom1234

Thank you, this solved the problem. But I got "OUT of Memory" message when it reached the "WanVideo Sampler" node , I increased the "blocks_to_swap" to 40, now it is somehow stuck at the sampler node. Is it possible to change the wan_animate models to gguf versions in your workflow? Or it needs an optimised Workflow for that?

Freno Deckilin

Thank you so much - that was the solution!

Freno Deckilin

got the same problem with a 4080. Restarting the system does not help. setting blocks to swap to 40 same problem. (video reference is 3 sec, 850x478px. picture is 1340x893px)

nurtu

Its nice that you have a .bat file but seriously it reinstalls already existing packages and fucks up my chain of compatible versions which results in no workflow no longer working. Im lucky i had a pip freeze so i can force reinstall my versions.

Victor Tabram

All works great!!!! best so far..... until I try the SEC masking. I get this error on SecModelLoader: Failed to load SeC model: No module named 'iopath' (I've got the model in the sams folder)

Victor Tabram

All good. Installing missing nodes via Comfy Manager didn't do the requirements! Had to run: ..\..\python_embeded\python.exe -m pip install -r ComfyUI-SecNodes\requirements.txt

nurtu

I need you to do a pip lsit for me so I know what Cuda, Torch and ptyhon you are running, i get fucked over every time you make an installer. Its easier if i just match your env

Reina Stryfe

Error No input found for flattened id [265] slot [-1] i get this error, how to fix?

Anthony

I am getting the error Prompt execution failed Prompt has no outputs:

bff udf

Prompt has no outputs

Aitrepreneur

I just changed the installer, it should now freeze your current env so less things should break

Aitrepreneur

dm me with more info, never saw this before

Aitrepreneur

you need to enable the workflow with the button on the left, check the video

Torsten

This doesend help. It looks like the emeded Python is part of the prolem. Python 3.13 seems not to work with CUDA?!?

Torsten

BTW the error only comes up when using SEC masking

Delavestra

Hi there! The rank64 lightx2v has versions 128 and rank 256... what's the difference? Do we want to use the latest ones?

Delavestra

What is the longest generation this wf can handle? Can it do 1-2 minutes, or capped at 5-10 seconds?

Jon Maddox

I have this same issue

IDE Busmaster

Also get cuda OOM on a 3090 and 48GB of RAM as well on a 60 frame clip at default size. Running a fresh comfy install from the AIO installer bat and the updated workflow.

Freno Deckilin

any other solutions yet on this error? @Aitrepreneur? (Thanks for help!!)

Melissa DeLaFey

Thank you as always Aitrepeneur! I hope you or someone here can help me out. When I run the workflow I get everything working except at the very end the final video is not generated/saved. I see all the process going without errors and in my output folder are only *.png and *.gif. The whole far right section of the two video combine nodes is not working. It is active but it does not do anything. How do you connect that section? I don't see any spaghetti going to that far right 3 nodes in your WF like they are isolated? Please send help! Cheers!

Freno Deckilin

@Torsten: i get the same error with or without SEC. i already tried with ultra low resolutions with the same problem

Victor Tabram

After using for a few days now this is the best animate WF by far. Amazing work!! However, MASKING video for a face replacement doesn't work very well and looks distorted and not much like the reference image, even with a good quality image. I've tried with face tracking nodes enabled and disabled, as per notes. What am I missing?

newtoai

When using a video to add the motion of the video to the picture, the output moving picture has the top part cut/the face cut , compared to the original picture

newtoai

Hi guys, is there a discord channel or a chat service to use to communicate to each other to help each other?

Sheel Doshi

hey! im wondering if its possible to replace the pose and face portion of this workflow with a more depth map like preprocessor. id like to use wan 2.2 to make things that were more similar to what QR monster controlnet used to be able to do using stable diffusion back in the day. is this possible?

EtD

Forever stuck at WanVideo Sampler node The display near the top left show: (1) 86% - WanVideo Sampler (#4 - 0%) I'm running on 4060 Ti 16gb VRAM It's been like this for about 45 minutes already, and Task Manager show ~0% GPU activity. Memory ram also at normal level (compared to when I don't use ComfyUI)

SataraN

I noticed this on my 3090 also, I'm thinking to try this with a lower python version and see if the same issue happens. *edited update - even with python on a lower version still seems to have the issue when the SeC nodes are used, the work flow works fine without the SeC but the moment you enable the SeC is when the issue takes place.

EtD

I've tried the following methods but it still doesn't work: - Fresh install everything. Using 1-Click install bat file - Manually delete "ComfyUI-WanVideoWrapper" folder in Custom Nodes. Then Install through ComfyUI manager UI, and also another try with URL - Cut down the frame cap to only 72

EtD

The terminal end with 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.

EtD

The 1-click install and workflow work fine on Runpod

Sean Rowe

Is there an option to create a image to video where we are generating a new video? I'd like to use this workflow with triton but use a reference image to generate a video (using loras) rather than modifying an existing video.

Melissa DeLaFey

Thank you as always Aitrepeneur! I hope you or someone here can help me out. When I run the workflow I get everything working except at the very end the final video is not generated/saved. I see all the process going without errors and in my output folder are only *.png and *.gif. The whole far right section of the two video combine nodes is not working. It is active but it does not do anything. How do you connect that section? I don't see any spaghetti going to that far right 3 nodes in your WF like they are isolated? Please send help! Cheers! EDIT: BTW I just want to transfer the movement of the video to the reference image, so I disabled the lower part of the WF.

TrapHouseMag

I'm having the same problem.

TrapHouseMag

What value do we put in the (Per_batch) box in the Resize Image v2 node?

Eyes_Wide_Shut

What Loras work with this? Wan 2.1 or Wan2.2?

red baron

i need a workflow to upscale 30 sec video. any idea? runpod also okay

ElektroMuse

I fixed this changing to the get mask module: mask and background_image

Glowbeard

Yeah same problem. But forcing the older front end like ElGalloUltimo said worked for me.

Mikael Svenson

If the reference image is not coming thru with good facial details in the video after replacing, what parameters should be adjusted to increase likeness? And an awesome workflow!

Mikael Svenson

Figured it out.. increased the face strength seems to help.

Bafangit

Thank you for this workflow. It's working perfectly on a 3090. I used your one-click bat for a completely clean install. This is so awesome after several weeks of unsuccessfully trying wan animate workflows only to have repeated errors. Looking forward to trying some of your other workflows. Keep up the fantastic work!

Giusparsifal

First of all, thanks! I have had always problem with GrowMask Node, as my reference charachter has long curly hair, after a lot of attempts I got that I had to grow from 10 to 70 to have right hair syle (otherwise my chachter will assume same haircut of the original charachter of the video). But since a couple of days this parameter is not good anymore, someone can explain me why? Something changed?

Melissa DeLaFey

Greetings! Now that I upgraded my membership I am re-posting hoping I can get help and make this work. I did a fresh install on a different drive using your *.bat and everything worked fine, even sage attention and triton. My problem persist and after the workflow finishes, I end up in my output folder with only 1 png for the face, 1 gif for the face, 1 png for the animation and 1 gif for the animation but NO VIDEO at all. The only difference from before is that now I SEE both Video Combination processes RAN but still the video is no where to be found and there's no video shown in the workflow itself. Please help! Thank you! FORGOT to mention that I have disabled the whole masking section 4 because I only want to transfer the animation from the reference video to the reference image or in other words, make the subject of my image do the movements from the reference video only while keeping the subject and background of my reference image.

ringlordwraith

I had the same issue. Its related to the resize ImageV2 node getting bad data from the 'get width' and 'get height' inputs. I was able to fix this by adding two new kjn integer constant nodes. rename one width and the other height. the width height values will be added to these nodes. ex. width 832 /height 480 for a landscape orientation. Plug these into the resize Imagv2 widht/height inputs (delete the get_width/get_height). Next plug in your new width and height nodes into set_width and set_height (constant nodes), and delete the old width and height that were inputs. Last thing plug in your new width height nodes into the Load Video node (where you add your video), by plugging into custom width and custom height inputs. That's it. That worked for me. I cant upload an image, but your new width heigh nodes will each have3 strings coming from them, one to set (width), one to custom(width) on Load video node, and one to (width input on the resize image v2 node. Hope that helps!

Melissa DeLaFey

@ringlordwraith Thank you for your suggestion! Im glad it worked for you. It didnt for me. i followed your suggestions but the workflow ends right after making the video (*.gif) of the animation and the farther right two video combines nodes of the third section dont do anything. Do you think you can try to share your workflow? Maybe I still did something wrong... Thanks

ringlordwraith

There is a similar workflow by MDMZ on Patreon . Out of respect for Aitrepreneur I wont post it here, but maybe give it a try. It doesnt have the advanced SEC masking, but it works fine.

Stefan Djokic

Have you resolved this? Even 40 steps doesn't work

Avishek

Which folder do the vitpose models like onnx models of whole body go into?

x:)

Anyone know how to add CacheDit Wan qccelerator to this workflow? Cant figure it out :(