⬢github Python · 8.7K ★ +89 since we first saw it · pushed 129 d ago · MIT
meituan-longcat/LongCat-Video
LongCat-Video is Meituan's open-source 13.6B-parameter video generation model. A single unified model handles text-to-video, image-to-video, and video-continuation, enabling minutes-long 720p/30fps videos without quality drift. It uses coarse-to-fine generation, block sparse attention for efficiency, and multi-reward GRPO RLHF, performing comparably to leading open-source and commercial models.
Why now: Recently released LongCat-Video-Avatar 1.5 (May 2026), keeping attention on the project alongside its strong open-source video generation benchmarks.
Who it is for: AI researchers and developers building or fine-tuning video generation pipelines who want a powerful open model.
Stars over our 24 snapshots: 8.6K to 8.7K, since 6 h ago.
Where people talked about it
- ⬢github meituan-longcat/LongCat-Video 6 min ago
API: https://socialmediatrends-api.osmike.com/v1/repos/meituan-longcat/LongCat-Video