V2SFlow: Video-to-Speech Generation with Speech Decomposition and Rectified Flow

TL;DR

V2SFlow introduces a novel framework that decomposes speech into subspaces and employs a rectified flow decoder to generate natural speech from silent videos, outperforming existing methods especially in unconstrained scenarios.

Contribution

The paper presents a new V2S framework using speech decomposition and rectified flow modeling, improving robustness and naturalness in real-world conditions.

Findings

Outperforms state-of-the-art V2S methods

Generates speech with higher naturalness than ground truth

Effective on unconstrained, real-world datasets

Abstract

In this paper, we introduce V2SFlow, a novel Video-to-Speech (V2S) framework designed to generate natural and intelligible speech directly from silent talking face videos. While recent V2S systems have shown promising results on constrained datasets with limited speakers and vocabularies, their performance often degrades on real-world, unconstrained datasets due to the inherent variability and complexity of speech signals. To address these challenges, we decompose the speech signal into manageable subspaces (content, pitch, and speaker information), each representing distinct speech attributes, and predict them directly from the visual input. To generate coherent and realistic speech from these predicted attributes, we employ a rectified flow matching decoder built on a Transformer architecture, which models efficient probabilistic pathways from random noise to the target speech distribution. Extensive experiments demonstrate that V2SFlow significantly outperforms state-of-the-art methods, even surpassing the naturalness of ground truth utterances. Code and models are available at: https://github.com/kaistmm/V2SFlow