FastVideo icon indicating copy to clipboard operation
FastVideo copied to clipboard
verified publisher icon hao-ai-lab

Video to video?

Open antithing opened this issue 7 months ago • 7 comments

Motivation

Hi, does this support video to video mode, with image guidance?

What are the hardware requirements?

Related resources

No response

antithing avatar May 28 '25 20:05 antithing

Currently not yet, I am assuming you are referring to models like VACE from wan?

SolitaryThinker avatar May 29 '25 17:05 SolitaryThinker

Yes, but directly via code, without using comfyui. Is this planned as a feature?

antithing avatar May 29 '25 20:05 antithing

This is planned, however no one is actively working on it as we are currently focused on supporting training in V1. Any contributions are welcome if you urgently want to see it supported

SolitaryThinker avatar May 29 '25 22:05 SolitaryThinker

Yes VACE support would be nice

bhack avatar Sep 16 '25 05:09 bhack

This is planned, however no one is actively working on it as we are currently focused on supporting training in V1. Any contributions are welcome if you urgently want to see it supported

Hello, I haven't seen any support for video-to-video at the moment. I was wondering if you have abandoned the plan for this feature?

KZF-kzf avatar Oct 03 '25 08:10 KZF-kzf

Coming soon! https://github.com/hao-ai-lab/FastVideo/pull/829

SolitaryThinker avatar Oct 03 '25 08:10 SolitaryThinker

Coming soon! #829

Very cool! But this seems to be only the inference code. Is the corresponding fine-tuning code for V2V also supported?😊

KZF-kzf avatar Oct 03 '25 09:10 KZF-kzf

I am also looking forward to the training code for vace lora😊

FastVideo is a great work by the way.

joe-zxh avatar Dec 15 '25 09:12 joe-zxh