Audio-Driven Universal Gaussian Head Avatars Journal: TOG Conference: SIGGRAPH Asia 2025 Conference Papers; December 15–18, 2025; Hong Kong, Hong Kong SIGGRAPH Asia 2025 Conference Papers (SA Conference Papers ’25), December 15–18, 2025, Hong Kong, Hong Kong
Kartik Teotia, Helge Rhodin, Mohit Mendiratta, Hyeongwoo Kim, Marc Habermann, Christian Theobalt · 2025
Read Audio-Driven Universal Gaussian Head Avatars Journal: TOG Conference: SIGGRAPH Asia 2025 Conference Papers; December 15–18, 2025; Hong Kong, Hong Kong SIGGRAPH Asia 2025 Conference Papers (SA Conference Papers ’25), December 15–18, 2025, Hong Kong, Hong Kong by Kartik Teotia, Helge Rhodin, Mohit Mendiratta, Hyeongwoo Kim, Marc Habermann, Christian Theobalt (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-18924.txt.
Description
Kartik Teotia and five co-authors present a way to build a moving, talking head avatar from sound alone. Past methods mostly turned speech into changes in head shape, and they missed how speech also changes how the face looks. Their system needs no prior model of the specific person. It pairs a general speech model with a learned prior for head avatars, trained on videos of many people seen from many angles.
The speech model turns raw audio straight into codes for facial expression. Those codes carry both shape and look, including eyebrow moves, gaze shifts, and the inside of the mouth. A light encoder then adapts the system to a new person from a single camera view. Tests show sharper lip sync and more lifelike images than shape-only methods.
Catalog record
| CREATOR | Kartik Teotia, Helge Rhodin, Mohit Mendiratta, Hyeongwoo Kim, Marc Habermann, Christian Theobalt |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-18924