A workflow using the latest wan2.2 fun VACE model to swap faces with high fidelity in a video.
Free try out link: https://www.runninghub.ai/post/1966792124551344129
Please check out my video for instructions before you start:
The video generation is resource-heavy. If your local computer is unable to run this workflow or you just want to see how good the model is and check all the workflow parameter settings before downloading, please open the link above to run it online for free on RTX4090. Just click and run. No brain-racking local setup.
1000 credits upon signing in using the above link(One gen only takes about 10-30 credits), with an extra 100 credits on daily login.
Description
FAQ
Comments (11)
I did get this to work, however, I had to hunt down a bunch of missing models that weren't listed. But anyway, I get a large black circle around the face after generation, how do I get rid of those?
Please adjust the erode and dilate values in the segment node to 1.
Thank you!
Man, this swaps so well! I'm sure everybody is lined up and waiting for a single frame swapper of this.
I know I do!)
Seriously, how can we modify this workflow to take image as an input instead of video, and outputting a png image with a swapped face?
Unfortunatelly not work for me
WanVideoModelLoader
'NoneType' object has no attribute 'endswith'
You need to download https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/blob/main/VACE/Wan2_2_Fun_VACE_module_A14B_HIGH_fp8_e4m3fn_scaled_KJ.safetensors and repoint the WANVideo VACE Module Select node.
CUDA error: out of memory
using all models in the workflow on a 5090. what gives ?
is there any workflow for swapping any element in a video with a given image?
Not working at all. Getting ghosts style strange artefakts. Could you doublecheck your workflow?
I have the same problem, did you find a solution ?
Seems difficult to dial in. After trying different LORAs and models, I am able to consistently get results that have graphical glitches over the subject's head. Sometimes their head becomes a big blurry blob; other times it looks like a mannequin version of the character I'm trying to swap in.
This workflow uses custom nodes you haven't installed yet.
Installation Required
Install RequiredWanVideoSetBlockSwap
Install RequiredWanVideoSetLoRAs
Install RequiredWanVideoVACEModelSelect
Install RequiredWanVideoDecode
Install RequiredWanVideoModelLoader
Install RequiredWanVideoSampler
Install RequiredWanVideoVAELoader
Install RequiredWanVideoLoraSelectMulti
Install RequiredWanVideoVACEEncode
Install RequiredLoadWanVideoT5TextEncoder
Install RequiredWanVideoTextEncode
Install RequiredWanVideoBlockSwap
Install RequiredInt
Install RequiredSeedVR2BlockSwap
Install RequiredSeedVR2
Install RequiredImpactInt
Install RequiredInpaintPreprocessor
Install RequiredLayerUtility: ImageScaleByAspectRatio V2
Install RequiredText Concatenate
Install RequiredLayerMask: HumanPartsUltra
Install Requiredeasy showAnything
Install RequiredImageScaleToMegapixels
Install RequiredRH_Captioner, trying from morning with chatgpt pro, still not solving, pls advice