Skip to content

URAvatar

Relightable head avatars from a phone scan

Visit official site

What URAvatar is

The technique is described as a way to create realistic relightable head avatars from a phone scan.

What you can do with it

  • Rebuild a head from a single-environment scan
  • Animate and relight the avatar in real time
  • Train a decoder with multi-view data

Getting started

  • Prepare a phone and a subject
  • Scan the subject under natural light
  • Reconstruct pose, shape and albedo from the scan

What to watch out for

  • A phone scan of a subject is described as input
  • A single-environment scan is the starting point
  • A pre-trained model is fine-tuned at the end

Pros & cons

✓ What we like

  • Training uses multi-view facial performance data
  • Personal fine-tuning through inverse rendering is listed

! What to watch out for

  • A phone scan of a subject is described as input
  • A single-environment scan is the starting point

FAQ

What is the input?

A phone scan of a subject is described.

How does it differ from earlier methods?

Radiance transfer learning is named instead of parameter estimation.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review