peer reviewed · WACV (2024)
Diff2Lip: Audio Conditioned Diffusion Models for Lip-Synchronization
Soumik Mukhopadhyay, et al.
In plain English
A lip-sync model that uses an audio-conditioned diffusion model to regenerate the mouth region while preserving identity, pose and image quality, tested on VoxCeleb2 and LRW.
lip sync
diffusion
talking head generation
evaluation