MeanFlow-TSE: One-Step Generative Target Speaker Extraction with Mean Flow

TL;DR

MeanFlow-TSE introduces a one-step generative framework for target speaker extraction that achieves high-quality separation in real-time, outperforming multi-step diffusion models in efficiency and accuracy.

Contribution

It presents a novel mean-flow-based one-step generative approach for TSE, eliminating the need for iterative sampling and enabling real-time application.

Findings

Outperforms existing diffusion- and flow-matching models in separation quality

Requires only a single inference step, enabling real-time processing

Demonstrates effectiveness on the Libri2Mix corpus

Abstract

Target speaker extraction (TSE) aims to isolate a desired speaker's voice from a multi-speaker mixture using auxiliary information such as a reference utterance. Although recent advances in diffusion and flow-matching models have improved TSE performance, these methods typically require multi-step sampling, which limits their practicality in low-latency settings. In this work, we propose MeanFlow-TSE, a one-step generative TSE framework trained with mean-flow objectives, enabling fast and high-quality generation without iterative refinement. Building on the AD-FlowTSE paradigm, our method defines a flow between the background and target source that is governed by the mixing ratio (MR). Experiments on the Libri2Mix corpus show that our approach outperforms existing diffusion- and flow-matching-based TSE models in separation quality and perceptual metrics while requiring only a single inference step. These results demonstrate that mean-flow-guided one-step generation offers an effective and efficient alternative for real-time target speaker extraction. Code is available at https://github.com/rikishimizu/MeanFlow-TSE.