signalGitHub Trending2026-10-05
meituan-longcat/LongCat-Video
Meituan released LongCat-Video, a 13.6B parameter foundational video generation model that unifies text-to-video, image-to-video, and video continuation tasks. It generates 720p 30fps videos efficiently using coarse-to-fine generation and block sparse attention, and achieves performance comparable to leading open-source and commercial models via multi-reward GRPO. The model is open-sourced with technical report and weights on Hugging Face.
- for who
- Video generation researchers and developers
- why now
- LongCat-Video-Avatar-1.5 released May 21, 2026, upgrades lip sync and long-video generation.
- what changes
- They can build long-form video generation systems without commercial APIs and experiment with a unified, high-quality open-source model.
- to do
- Clone the GitHub repo, set up the environment, and download the Hugging Face weights to run LongCat-Video.
key points
- 13.6B parameters, unifies text-to-video, image-to-video, and video continuation
- Generates 720p 30fps long videos without quality degradation using coarse-to-fine and block sparse attention
- Comparable to leading open-source and commercial models via multi-reward GRPO
#video generation#open source model#LongCat-Video#Meituan
score
score 7 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source