Hugging Face's logo

lightx2v
/
Wan2.2-Lightning

Text-to-Video
Safetensors
English
text-to-video;
image-to-video;
comfyUI;
video-generation;
Model card Files Community
29

I2V weights released with kijai-wrapper-compatible workflow. T2V V1.1 released.

#16
by X-niper - opened
Discussion
X-niper
•
edited 6 days ago

Please take a try with latest kijai's wan video wrapper version.

Welcome to reproduce the results in the demo cases using our repo https://github.com/ModelTC/Wan2.2-Lightning.

I2V workflow: https://huggingface.co/lightx2v/Wan2.2-Lightning/blob/main/Wan2.2-I2V-A14B-4steps-lora-rank64-Seko-V1/Wan2.2-I2V-A14B-4steps-lora-rank64-Seko-V1-forKJ.json

T2V workflow: https://huggingface.co/lightx2v/Wan2.2-Lightning/blob/main/Wan2.2-T2V-A14B-4steps-lora-rank64-Seko-V1.1/Wan2.2-T2V-A14B-4steps-lora-rank64-Seko-V1.1-forKJ.json

tonynoce

https://files.catbox.moe/1mw30j.mp4

euler / beta

Nice result I got there, a bit less motion but I like the version with lora

wywywywy

In the workflows, the shift parameter is set to 1.0. Is this expected or is it a mistake?

In the Github code it seems to default to 5.0.

X-niper
•
edited 6 days ago

@wywywywy It's expected because we explicitly pass in the sigma list.

wywywywy

@wywywywy It's expected because we explicitly pass in the sigma list.

I see. Thank you. Hopefully we can soon do that in native workflows as well.

tonynoce

workflow (1).png

Oh so @X-niper the modelsamplingSD3 should be on 1 ? I used it on 8...

X-niper

@tonynoce

Hi, for the native comfy workflow, it should be 5. We have released the template workflow for both T2V and I2V. Please refer to the discussion. The workflow file that contains "forKJ" is for kijai's wrapper and that contains "NativeComfy" is for native comfyUI.

Sign up or log in to comment