Skip to content

DICE-Talk

Generate talking portraits with varied emotions

Visit official site

What DICE-Talk is

The method is a diffusion-based talking-portrait technique that decouples identity and emotion.

What you can do with it

  • Generate an emotional talking portrait
  • Feed an image and an audio file
  • Pick the emotion to render

Getting started

  • Install the required libraries
  • Download the model files
  • Run the demo with the input paths

Pros & cons

✓ What we like

  • A demo and graphical interface are provided
  • A high-performance GPU speeds up generation

! What to watch out for

  • Software libraries must be installed first
  • Model files are downloaded before use

FAQ

Which emotions are listed?

Neutral, happy, angry and surprised.

What inputs are supported?

Images and audio are accepted.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review