We are thrilled to announce the open-source release of UniVideo, the first model to successfully unify video understanding, generation, and editing into a single framework.
The Team
This project is a collaboration between researchers at Kuaishou Technology and the University of Waterloo.
Key Contributors:
Jiabin Luo, Junhui Lin, Zeyu Zhang, Biao Wu, Meng Fang, Ling Chen, Hao Tang, and Wenhu Chen.
Availability
The code and model weights were officially released on January 7, 2026. You can now access the repositories and start building.
Future Roadmap
While today marks a significant milestone, we are already working on v2. The next iteration will focus on prolonged video consistency (up to 2 minutes) and real-time audio synchronization.
Stay tuned for updates and join our community to shape the future of generative video.