FlashLips: 100-FPS Mask-Free Latent Lip-Sync using Reconstruction Instead of Diffusion or GANs

TL;DR

FlashLips is a fast, mask-free lip-sync system that uses reconstruction and self-supervision to achieve real-time performance with high visual quality, avoiding GANs or diffusion models.

Contribution

It introduces a two-stage pipeline with a latent-space editor and an audio-to-pose transformer, enabling 100 FPS lip-sync without explicit masks or generative adversarial training.

Findings

Achieves over 100 FPS on a single GPU.

Matches the visual quality of larger state-of-the-art models.

Uses self-supervision to localize lip edits without explicit masks.

Abstract

We present FlashLips, a two-stage, mask-free lip-sync system that decouples lips control from rendering and achieves real-time performance, with our U-Net variant running at over 100 FPS on a single GPU, while matching the visual quality of larger state-of-the-art models. Stage 1 is a compact, one-step latent-space editor that reconstructs an image using a reference identity, a masked target frame, and a low-dimensional lips-pose vector, trained purely with reconstruction losses - no GANs or diffusion. To remove explicit masks at inference, we use self-supervision via mouth-altered target variants as pseudo ground truth, teaching the network to localize lip edits while preserving the rest. Stage 2 is an audio-to-pose transformer trained with a flow-matching objective to predict lips-pose vectors from speech. Together, these stages form a simple and stable pipeline that combines deterministic reconstruction with robust audio control, delivering high perceptual quality and faster-than-real-time speed.