Meituan has open-sourced version 1.5 of its LongCat-Video-Avatar framework, which generates photorealistic digital human videos. This updated framework achieves state-of-the-art lip-sync accuracy using only eight inference steps.
Meituan’s release makes a photorealistic digital-human video framework available in open source form, with lip-sync accuracy achieved in just 8 inference steps. That matters for streaming and video tooling because it lowers the compute burden of producing avatar-based video, while improving one of the most visible failure modes in synthetic presenters. The key signal to watch next is whether Meituan publishes additional benchmarks or demos beyond the lip-sync result for LongCat-Video-Avatar 1.5.
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source