Loki: Representation over Architecture for Diffusion-Based Portrait Animation
Pouyan Navard, Sernam Lim · 2026
Read Loki: Representation over Architecture for Diffusion-Based Portrait Animation by Pouyan Navard, Sernam Lim (2026) 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-2605-24176.txt.
Description
Navard and Lim present Loki, a portrait-animation system that moves a driver clip's expression and pose onto one reference photo, keeping its identity. Leading diffusion systems stack modules for expression, pose, and identity, paying in parameters and tangled controls. Loki instead encodes driver expression and pose with a face model whose axes stand orthogonal to identity, then maps them for the backbone to read natively. Identity flows separately through backbone features. Cross-identity reenactment then needs a coefficient swap, and Loki uses about 43 percent fewer parameters than rivals.
Catalog record
| CREATOR | Pouyan Navard, Sernam Lim |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 86506 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.24176