Skip to content

ExAvatar

Build a 3D full-body expressive avatar from a phone scan

Visit official site

What ExAvatar is

The model is described as a 3D full-body dynamic expression model that pairs the body-driving capability of SMPL-X with the appearance modelling of 3DGS, and it can be created from a simple phone scan.

What you can do with it

  • Drive the body, hands and face as a whole
  • Keep the hybrid representation compatible with the SMPL-X expression space
  • Reduce artefacts on new expressions and poses with a connection-based regulariser
  • Lift hand and face expressiveness through joint and face offsets

Getting started

  • Prepare a phone and a subject
  • Record a neutral pose for about 10 seconds
  • Upload the recorded video to the system
  • Let the system recognise the subject and build the 3D model

Example uses

  • A 3D model demo of the Standing Next To You dance
  • The Supernova dance shown as a 3D demo
  • The Smooth Criminal dance rendered on a full-body avatar

Pros & cons

✓ What we like

  • A short phone scan is enough to create the model
  • Joint and face offsets raise hand and face expressiveness

! What to watch out for

  • The workflow starts from a neutral-pose phone video
  • Animation runs through LBS and rendering relies on 3DGS

FAQ

How is the avatar created?

The steps list a phone, a subject, a roughly 10-second neutral-pose clip, an upload and automatic 3D model creation.

What does the hybrid representation change?

The material states it improves facial naturalness, reduces artefacts and stays compatible with the SMPL-X expression space.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review