VACE
VACE copied to clipboard
ali-vilab
Official implementations for paper: VACE: All-in-One Video Creation and Editing
Thank you so much for doing such a great job，I am working on person-swap now, but I found that the performance of vace-1.3b is not so good, and the performance...
The temporal extension is neat on paper, but I'm getting a shift in color that makes it obvious where the extension occurs. Is this a limitation or is there some...
- fix: add 'device=None' in annotators' constructors for schema consistency and compatibility - typo: gpu instead of cpu when cuda is_available - fix: for when device is provided as string
Hi everyone, VACE is... WONDERFUL! Looking to go beyond 720p, and knowing that Wan Video already supports 1080p, is there a planned version of VACE that will support 1080p? Thanks...
May I know if it is possible to transfer the motion from the reference video onto the subject in the reference image ? How should the VCU be set ?
Hi ! Maybe a silly question. What's the relationship between the VACE version and the original version ? How should we choose ?
The editing capabilities are great, it would be great if the camera editing could be customized. Currently, Wan2.1 has some models for controlling the lens, but they don't seem to...
Thank you for your excellent work! We noticed that in the illustrative figure for the Masked Video-to-Video Editing task, subject removal (inpainting) is demonstrated. However, we couldn't find a concrete...
Hi VACE team, I'm trying to reproduce the training script; however, I've encountered some issues. I'm using the LTX-Video base model to train on face generation (512x512 resolution) with these...
