LongCat-Video (Meituan)

▶ See it at the source

What they did

A 13.6B open video model that handles text-to-video, image-to-video and video continuation in one network. It is pretrained on continuation, so it can produce minutes-long videos without colour drift, and it renders 720p at 30 fps using a coarse-to-fine strategy and block-sparse attention.

Worth knowing

• Price: Open weights on Hugging Face (mit)
• License: mit
• Status: sourced. We have not run this ourselves yet (no "Tested" badge).
• Quality bar (computed by code): 4/5 (primary source ✓, reproducible ✗, new ✓, example output ✓, safe ✓)