Unified Character Video Editing for Live Streaming

Zhiyuan Li^1,3 · Chi-Man Pun^1,📪 · Peng-Tao Jiang^2,📪 · Bo Li² · Xiaodong Cun^3,🚩

¹ University of Macau    ² vivo BlueImage Lab    ³ GVC Lab, Great Bay University

^📪 Corresponding authors    ^🚩 Project lead

EditaLive_showreel.mp4

📋 TODO

• If you find EditaLive useful or interesting, please give us a Star 🌟. Your support encourages us to keep improving the project.
• We are doing our best to organize and polish the code for public release 🏃💨. Thank you for your interest and patience 🙏.
• Release training code.
• Release inference code, config, and pretrained weights.
• [2026.08.27] 🔥 Release paper.

⚖️ Disclaimer

• This project is released for academic research only.
• Users must not use this repository to generate harmful, defamatory, or illegal content.
• The authors bear no responsibility for any misuse or legal consequences arising from the use of this tool.
• By using this code, you agree that you are solely responsible for any content generated.

🚀 Release Status

The code, pretrained weights, and detailed installation and inference instructions are being prepared for public release. Please watch this repository for updates.

⚙️ Framework

❤️ Acknowledgement

This repository is mainly built upon Wan-Animate, VideoX-Fun, LightX2V, and FastVideo, thanks to their invaluable contributions.