-
[6/30/2024] After half a year of negotiations, I regret to inform you that we are unable to release the FlowVid checkpoints. I understand this is disappointing to the community, and I sincerely apologize to let you down. I recommend checking out TokenFlow, which has on-par results as our FlowVid, and our StreamV2V, the fastest video-to-video model available. Although I cannot release the checkpoints, I am permitted to share code snippets upon request. If you are interested in exploring the implementations, please contact me at [email protected], and I will guide you to the code.
-
[12/28/2023] The official codes and models are currently undergoing a legal review process. Please stay updated for further developments. If you are interested in re-implementing FlowVid or require specific code snippets, don't hesitate to contact me at [email protected]. Thanks!
FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis
Feng Liang,
Bichen Wu,
Jialiang Wang,
Licheng Yu,
Kunpeng Li,
Yinan Zhao,
Ishan Misra,
Jia-Bin Huang,
Peizhao Zhang,
Peter Vajda,
Diana Marculescu
- [12/28/2023] The official codes and models are currently undergoing a legal review process. Please stay updated for further developments. If you are interested in re-implementing FlowVid or require specific code snippets, don't hesitate to contact me at [email protected]. Thanks!
This codebase builds on diffusers. Thanks for open-sourcing!