DiffSwap++: 3D Latent-Controlled Diffusion for Identity-Preserving Face Swapping

Weston Bondurant Arkaprava Sinha Hieu Le Srijan Das Stephanie Schuckers UNC Charlotte {wbondura,sdas24}@charlotte.edu · 2025

Read DiffSwap++: 3D Latent-Controlled Diffusion for Identity-Preserving Face Swapping by Weston Bondurant Arkaprava Sinha Hieu Le Srijan Das Stephanie Schuckers UNC Charlotte {wbondura,sdas24}@charlotte.edu (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-2511-05575.txt.

Description

Weston Bondurant and co-authors present DiffSwap++, a system that swaps one person's face onto another's in images and video. Diffusion methods beat older approaches on image quality, yet they still smudge fine detail and lose identity, above all with hard poses or expressions. A core flaw is that past systems barely use three-dimensional face structure.

DiffSwap++ feeds 3D facial features into training, which helps split identity from pose and expression. The design also conditions each denoising step on identity codes and facial landmarks. On three public face data sets it beat prior methods at keeping the source identity while holding the target pose. A biometric-style test and a user study backed the gains.

Catalog record

CREATORWeston Bondurant Arkaprava Sinha Hieu Le Srijan Das Stephanie Schuckers UNC Charlotte {wbondura,sdas24}@charlotte.edu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-05575

Browse the full library · About the Tower of Babel