Skip to content

JoyVASA

Audio-driven portrait animation with diffusion

Visit official site

What JoyVASA is

The technique is described as a diffusion-based method that splits dynamic expression from static 3D face representation to produce facial motion and head movement.

What you can do with it

  • Generate longer portrait videos
  • Sample motion sequences from audio
  • Render animation from 3D keypoints
  • Extend the method to animal faces

Getting started

  • Extract 3D face features from a reference image
  • Extract audio features from the input speech
  • Sample motion with a sliding window

Pros & cons

✓ What we like

  • Longer video generation is listed
  • A mixed Chinese and English dataset is used

! What to watch out for

  • It is built on diffusion models
  • A reference image is needed as input

FAQ

What drives the animation?

The material describes audio-driven motion generation.

Does it handle animals?

The material states animal face animation is supported.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review