LongCat Avatar 1.6 AI Video Generator
LongCat Avatar 1.6 turns recorded audio into long-form avatar videos with natural lip sync, expressive facial and body motion, and consistent character appearance.
The generator below currently runs Avatalk 1.5.
Click to upload image
JPG, PNG, WEBP. Max 10MB
Click to upload audio
Audio max 30s for billing. MP3, WAV, M4A, OGG, FLAC.
Avatalk 1.5 AI Avatar Generator demo preview
What Is LongCat Avatar 1.6?
LongCat Avatar 1.6 is Meituan’s open-source model for turning audio and visual inputs into animated character video.
A reference image gives the model a subject and a starting look. The audio track gives it the timing for speech or singing. A text prompt can describe the scene or delivery. The official release also documents single-person generation from audio and a text prompt and a separate workflow for continuing a generated video.
LongCat Avatar 1.6 is short for LongCat-Video-Avatar 1.6. The project publishes model weights and inference instructions, along with examples of a presenter speaking, a singer performing, and characters exchanging dialogue. The samples let you hear how the mouth movements line up with the audio and see what happens to the face and framing over time.
LongCat Avatar 1.6 Video Demos
Play the original examples with sound and use the timeline to inspect specific scenes or compare the two versions.
Outdoor Talking Avatar
Duration 10:05Watch the speaker’s face and gestures beside the car.
Indoor Presenter
Duration 10:05Compare the speaker’s expression and framing near the start and finish.
Stage Singing
Duration 0:12Listen for sustained notes and watch the singer’s timing.
Two-Person Dialogue
Duration 0:10Watch the character in view as the conversation shifts.
Side-by-Side Close-Up
Duration 0:19LongCat Video Avatar 1.5 is on the left and 1.6 is on the right.
Side-by-Side Portrait
Duration 0:14Listen to the same line while comparing mouth shapes and expression.
How It Works
Create a single-avatar or two-speaker video with Avatalk 1.5.
Upload an image and audio
Sign in and choose Single Avatar or Multi-Person. Upload a reference image and your recording. For two speakers, assign each audio track to the person on the left or right.
Choose your settings
Select 480p or 720p and add a prompt if you want to describe the shot or expression. Check the credit cost before you submit.
Generate and download
Generate the clip and play it with sound to check the timing. Download the video from the preview panel when you’re happy with it.
LongCat Avatar 1.6 Features
The 1.6 release focuses on speech timing, coordinated motion, and keeping the same character recognizable through longer shots.
Lip-Sync Video
LongCat Avatar 1.6 aims to make mouth movements less exaggerated and more consistent from frame to frame. Listen to the voice while watching the lips at the start and end of a phrase. Pauses and changes in pace can make timing differences easier to spot.
Facial Expression and Body Motion
The model update also targets how expressions and body movement fit together. In the wider shots, look at the head, shoulders, and hands while listening to the delivery. The singing and dialogue demos give you different kinds of movement to judge.
Long-Form Avatar Video
LongCat Avatar 1.6 uses video continuation to build a longer sequence in segments. The project shows two clips just over ten minutes. Scrub from the first minute to the final minute to see how the face, clothing, and background hold up across each shot.
Single- and Multi-Person Scenes
The official instructions cover a single character and multi-person animation from an image. The multi-person workflow uses audio for the characters in the scene. Watch the listener during a line of dialogue to see how the model handles a face that is in frame while someone else speaks.
Faster Model Inference
Meituan documents a faster generation pipeline in LongCat Avatar 1.6, including settings that reduce memory use on supported hardware. Render time depends on the hardware and settings you choose. The quick-start instructions explain which options can be used together.
What Can You Create with LongCat Avatar 1.6?
Use spoken or sung audio to animate a solo performer, a presenter, or a scene with multiple people.
Talking Avatar
Pair a portrait with narration for a close-up speaker. This format gives viewers a clear view of the face as the delivery changes pace.
Long Presentation
Use the continuation workflow for a speaker delivering a longer recording in the same setting. The full-length demos show what that looks like in an indoor and an outdoor shot.
Singing Performance
Start with a vocal track and a reference frame. In the stage demo, the mouth and body have to follow sustained notes as well as quickly changing phrases.
Character Dialogue
Use a scene with more than one person and audio for the speakers. The dialogue example gives you a short scene where you can watch for changes in who is speaking.
Who Is LongCat Avatar 1.6 For?
The research demos are useful if your project depends on visible speech, consistent characters, or more than one voice.
Educators Recording Explanations
Use the presenter examples to judge whether a talking avatar fits a narrated lesson. Check how it looks through a quiet pause and a run of fast speech. These are two points where timing is easier to assess than in a still image.
Video Creators Working with Voice and Music
If you build scenes around recorded dialogue or vocals, watch the short LongCat Avatar 1.6 demos with the sound on. The performance and dialogue clips show different challenges, from stretching a vowel across a note to keeping a listener in frame.
Developers Evaluating an Open Model
Download the LongCat Avatar 1.6 weights and use the official examples to check the setup on your machine. Start with a short input so you can check the face and voice alignment before committing GPU time to longer sequences.
Choose Your Avatalk Credit Pack
Choose a one-time credit pack for the Avatalk 1.5 generator on this page. It offers 480p and 720p and shows the credit cost before you submit.
Base
Pro
Ultimate
Creator
Choose one-time credits • Flexible billing options
LongCat Avatar 1.6 FAQ
Get answers about inputs, video length, two-speaker scenes, and running the model yourself.
Can I turn a photo and a recording into a talking avatar?
Yes, you can animate a still photo with a voice recording using the Avatalk 1.5 generator on this page. LongCat Avatar 1.6 is currently unavailable in the online generator.
Can I start with a prompt without uploading a photo?
When you run LongCat Avatar 1.6 locally, you can generate a single-speaker video from a text prompt and a recording. The prompt describes the character and scene, and the recording provides the speech. This page’s Avatalk 1.5 generator requires a reference image.
How long can LongCat Avatar 1.6 extend a talking shot?
LongCat Avatar 1.6 can generate videos longer than 10 minutes by continuing the shot in segments. Meituan’s published specifications leave the maximum duration unspecified.
Can I animate two speakers in one video?
Yes. Start with a reference image showing both people and a separate recording for each speaker. In the online Avatalk 1.5 generator, choose Multi-Person and assign each recording to the person on the left or right.
What’s new in LongCat Avatar 1.6 compared with 1.5?
The 1.6 model improves lip movement, coordinates facial expressions and body motion more naturally, and keeps the character’s appearance more stable in longer shots. Meituan documents eight-step generation workflows for both versions.
Is the LongCat Avatar 1.6 model open source?
Yes. Meituan releases the LongCat Avatar 1.6 weights under the MIT License. Running it yourself requires the LongCat-Video base weights and the separate 1.6 avatar checkpoint.
Explore LongCat Avatar 1.6
Watch the model demos or make a clip online with Avatalk 1.5.