LongCat Avatar

The free, open-source talking-avatar model. One photo + one audio clip = a lip-synced talking video.

Added July 15, 2026

#ai-avatar#open-source#lip-sync#meituan#free-tools

Built and open-sourced by Meituan's LongCat team under the MIT licence, so the model itself is free — no subscription and no per-minute billing when you run it yourself.

Part 1 — What it can do

The core feature

Photo + audio to talking video

Upload a single portrait and an audio clip. LongCat gives you back a proper talking avatar, with mouth movement that matches the audio frame for frame.

Whisper-Large audio encoder

Accurate lip sync

Version 1.5 upgraded the audio encoder to Whisper-Large, which reads phonemes and rhythm more finely — so mouth shapes land on the right sounds.

Not just speech

Singing avatars

Feed it a song and the avatar performs it — rhythm-aware head and body motion, not a static face with a moving mouth.

Natural movement, not a puppet

Full-body avatars

Gestures, posture shifts and hand movement stay coordinated with the audio, so the body reads as a person actually speaking.

No freeze between lines

Natural during silence

Most avatar tools go stiff the second the audio stops. LongCat separates speech from motion, so blinking, breathing and small posture shifts carry on through pauses.

Multi-person support

Multiple characters in one scene

Two or more speakers in the same shot, each holding their own identity, with natural speaker/listener turn-taking.

Part 2 — TWO WAYS TO ACCESS IT

Option 1 - Hugging Face

THE FREE ROUTE

Try it straight in your browser on a community-hosted Hugging Face Space: upload a portrait and an audio clip, nothing to install, no card needed. Free, but you share a queue and clip length is capped. Want it uncapped? The model weights are on Hugging Face under the MIT licence - download them and run LongCat on your own GPU for free. https://huggingface.co/spaces/victor/LongCat-Video-Avatar-1.5

Option 2 - LongCat Avatar web app

THE NO-SETUP ROUTE (PAID CREDITS)

A hosted website where everything runs in the browser - no install, no GPU needed. Worth knowing before you click: this site is run by a third party, not by Meituan, and it charges credits per second of video. Convenient, but not the free route. The model underneath is the same one you can self-host for nothing. https://www.longcatai.net/longcat-avatar

Official code + install guide

GITHUB - MEITUAN LONGCAT TEAM

Setup steps, weight downloads and the exact inference commands. This is the route if you want LongCat running locally with no queue, no cap and no credits. https://github.com/meituan-longcat/LongCat-Video

Official model weights

HUGGING FACE - MIT LICENCE

The model card for LongCat-Video-Avatar 1.5: licence terms, benchmarks and the technical detail behind every feature on page 1. https://huggingface.co/meituan-longcat/LongCat-Video-Avatar-1.5

BEFORE YOU CLOSE THIS

Save this URL - open-source pages move and links change fast.
Share it with the friend who is still paying monthly for avatar videos

Disclaimer. This guide is for educational and informational purposes only. All brand, product and model names belong to their respective owners. WE are not affiliated with, endorsed by or sponsored by Meituan, the LongCat team, Hugging Face or any other brand featured here, unless explicitly stated. Tool availability, pricing, free tiers, licences and features can change at any time — always check the official source before you rely on them.