Tiny is not small enough: High-quality, low-resource facial animation models through hybrid knowledge distillation

A studio title card: white headline type floating over a looping swirl of magenta and violet wire, then a quiet gray paper underneath.

electronicarts.github.io/tiny-voice2faceTiny is not small enough: High-quality, low-resource facial animation models through hybrid knowledge distillation
Screenshot of https://electronicarts.github.io/tiny-voice2face/, © the authors. Shown for reference. Request removal
Venue
SIGGRAPH 2025
Year
2025
First author
Zhen Han
Field
facial animation
Template
Academic Project Page Template (Nerfies, Bulma 0.9) with custom video cover
Theme
Light

Checked against the rendered page. Real-time demo video renders as a black frame (no autoplay attribute, empty poster). Cover height is 60vh, so the full-page shot shows a taller cover than the 1440x900 render.

Why it is here

Full-bleed generative-art hero with a centred logo and title. It is the only cinematic, poster-like opener in the pool, and the clean centred metadata below it follows through. One flaw is that the demo video was captured as a black frame.

Overview

Color Palette

Typography

Google Sans
Tiny is not small enough: High-quality, low-resource facial animation Abstract · Method · Results · BibTeX — 0123456789
Noto Sans
Tiny is not small enough: High-quality, low-resource facial animation Abstract · Method · Results · BibTeX — 0123456789

Spacing & Shape

Page Modules

Guidelines

More like this