Skip to content

EmoTalk3D

Emotion-controllable 3D head synthesis from speech

Visit official site

What EmoTalk3D is

The project is described as research on 3D virtual head synthesis that uses multi-view video, emotion annotations and per-frame 3D geometry to address view consistency and emotional expression.

What you can do with it

  • Synthesise 3D heads with controllable emotion
  • Capture wide viewing angles in animation
  • Render dynamic facial details such as wrinkles

Model notes

  • An encoder separates content and emotion features in speech
  • S2GNet predicts dynamic 3D point clouds
  • A rendering module produces free-view animation

Pros & cons

✓ What we like

  • Wide viewing angles and dynamic facial details are described
  • Lip sync and rendering quality are stated as enhanced

! What to watch out for

  • The work is described as a research project
  • Input is prepared as audio with the needed emotion

FAQ

What does S2GNet do?

The material states it predicts dynamic 3D point clouds from speech.

What does the dataset contain?

The material states multi-view head data with emotion annotations.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review