Skip to main content

A free, open source Meituan model for audio-driven talking avatar videos, including clips longer than 10 minutes.

About LongCat Video Avatar 1.6

LongCat Video Avatar 1.6 is a free, open source model from the Meituan LongCat team for audio-driven human video generation. The official model is on Hugging Face at huggingface.co, and there is a project page at meigen-ai.github.io. The model card on Hugging Face was updated on November 13, 2026.

The card describes 1.6 as an upgrade from LongCat-Video-Avatar 1.5. Meituan says it improves expressive naturalness, lip and audio synchronization, coordinated facial and body motion, and visual consistency in long-form generation, and that it reduces exaggerated or strained mouth movements. It also says the model mitigates error accumulation during autoregressive video continuation, so it can generate videos longer than 10 minutes while keeping visual quality, identity consistency and temporal coherence.

It supports speech-to-video from text or from a reference image, plus video continuation, for a single person or several people. The card lists English and Chinese, and it is built on the LongCat-Video base model. The model weights are released under the MIT License. Installation uses the LongCat-Video GitHub repository, with optional INT8 or FP8 quantization. The card says the model has not been comprehensively evaluated for every downstream use, and that developers should assess accuracy, safety and fairness before deploying it in sensitive settings.

Pricing: Free. The model weights are released under the MIT License. Checked November 16, 2026.

Pricing model: Free

Features

  • Audio-driven talking avatar video generation
  • Single-person and multi-person animation
  • Speech-to-video from text or from a reference image
  • Video continuation for clips the model card says can exceed 10 minutes
  • English and Chinese
  • MIT license for the model weights
  • Weights and code on Hugging Face, with a project page and ModelScope copy

FAQs

Common questions about LongCat Video Avatar 1.6

What is LongCat Video Avatar 1.6?
A free, open source Meituan model for audio-driven talking avatar videos, including single-person and multi-person generation and clips the model card says can exceed 10 minutes.
Is it free to use?
Yes. The model weights are released under the MIT License. That license does not grant rights to use Meituan trademarks or patents.
Where do I get it?
The official model is on Hugging Face. There is also a project page and a ModelScope copy, and the code lives in the LongCat-Video GitHub repository.
How long can the videos be?
The model card says version 1.6 can generate videos longer than 10 minutes, using video continuation while trying to keep visual quality and identity consistent.

Featured