Skip to content

One Shot, One Talk

A full-body talking avatar from one photo

Visit official site

What one-shot-one-talk is

The technique is a deep-learning image generation method that reconstructs a personalized full-body talking avatar from one picture, with animated body movement and expression.

What you can do with it

  • Rebuild personalized details from a single photo
  • Produce animation with body and face changes
  • Use pseudo-label frames for dynamic modeling

Getting started

  • Visit the project page and download the code
  • Prepare one personal picture as input
  • Configure the environment and run the code

Pros & cons

✓ What we like

  • One photo is enough as the input
  • Generated motion covers both body and face

! What to watch out for

  • Code and environment setup are needed before running
  • Pseudo-labels come from pretrained generative models

FAQ

What is the input?

The material says a single personal image is prepared as the input.

Who developed it?

The material names researchers from the University of Science and Technology of China and Hong Kong Polytechnic University.

Last reviewed: 2026-11-15

More AI digital human tools

View all →

How we review