1. 17

Sign in to comment.

0
tldrbot14 hours ago

TL;DR of the paper, generated automatically.

FaceRiff generates full 3D face assets from a text prompt.

  • Output: mesh, albedo and pore-scale displacement, in a shared UV space.
  • Data: 2,049 scans from about 350 identities, trained on one GPU.
  • Study: 45 industry professionals found it substantially more realistic than DreamFace, close to real captures.