StuPASE: Towards Low-Hallucination Studio-Quality Generative Speech Enhancement

TL;DR

StuPASE, built upon PASE, is proposed, to achieve studio-level quality while retaining its low-hallucination property, and to address performance limitations under strong additive noise, is replaced with a flow-matching module.

Abstract

Achieving high perceptual quality without hallucination remains a challenge in generative speech enhancement (SE). A representative approach, PASE, is robust to hallucination but has limited perceptual quality under adverse conditions. We propose StuPASE, built upon PASE to achieve studio-level quality while retaining its low-hallucination property. First, we show that finetuning PASE with dry targets rather than targets containing simulated early reflections substantially improves dereverberation. Second, to address performance limitations under strong additive noise, we replace the GAN-based generative module in PASE with a flow-matching module, enabling studio-quality generation even under highly challenging conditions. Experiments demonstrate that StuPASE consistently produces perceptually high-quality speech while maintaining low hallucination, outperforming state-of-the-art SE methods. Audio demos are available at: https://xiaobin-rong.github.io/stupase_demo/.

Figures (2)

• Figure 1: Overview of the proposed StuPASE framework.
• Figure 2: Subjective evaluation results for perceptual quality (Q-MOS) and speaker similarity (S-MOS).