█

AI/TLDR

Meituan LongCat · 2026-11-13 · notable

LongCat-Video-Avatar 1.6 — Meituan's open avatar model for long talking videos

In short

LongCat-Video-Avatar 1.6 is Meituan's open audio-driven avatar model. It turns speech plus a photo or text prompt into talking video, with more natural lip sync, less drift in videos over 10 minutes and faster inference than 1.5.

Meituan LongCat · 2026-11-13 · 3 sources checked

Hugging Face model card preview for meituan-longcat/LongCat-Video-Avatar-1.6

Meituan's open avatar model makes people talk, sing and move in sync with audio, now for videos over 10 minutes.

What is it?

LongCat-Video-Avatar 1.6 is an upgrade of Meituan's open avatar model built on LongCat-Video. Given a speech track and either a reference image or a text prompt, it generates a video of one or several people speaking, with mouth, face and body motion matched to the audio.

How does it work?

Version 1.6 keeps the 1.5 base architecture and adds two LoRA adapters (DMD and OPSD) plus an optional INT8 DiT for faster inference. Long videos are built by audio-driven continuation in segments, and the release reduces the error that builds up across those segments, so faces and lips stay stable in clips longer than 10 minutes.

Why does it matter?

Most strong talking-avatar systems, such as OmniHuman and HeyGen, are closed services. Meituan's model card reports that in a human preference study 1.6 ranks second of six systems with an Elo of 1068.1, behind OmniHuman-2.0 and ahead of its own 1.5, while the weights are published under the MIT license.

Who is it for?

video and avatar developers, digital-human and dubbing teams

Try it

huggingface-cli download meituan-longcat/LongCat-Video-Avatar-1.6 --local-dir ./weights/LongCat-Video-Avatar-1.6

Sources · 3 outlets

Tags

  • meituan
  • longcat
  • longcat-video
  • video-generation
  • talking-avatar
  • lip-sync
  • audio-to-video
  • open-weights
  • mit
  • lora

← All releases · Learn AI