Accelerating Flow-Matching-Based Text-to-Speech via Empirically Pruned Step Sampling

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zheng, Qixi, Chen, Yushen, Niu, Zhikang, Ma, Ziyang, Wang, Xiaofei, Yu, Kai, Chen, Xie
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913874252398592
author Zheng, Qixi
Chen, Yushen
Niu, Zhikang
Ma, Ziyang
Wang, Xiaofei
Yu, Kai
Chen, Xie
author_facet Zheng, Qixi
Chen, Yushen
Niu, Zhikang
Ma, Ziyang
Wang, Xiaofei
Yu, Kai
Chen, Xie
contents Flow-matching-based text-to-speech (TTS) models, such as Voicebox, E2 TTS, and F5-TTS, have attracted significant attention in recent years. These models require multiple sampling steps to reconstruct speech from noise, making inference speed a key challenge. Reducing the number of sampling steps can greatly improve inference efficiency. To this end, we introduce Fast F5-TTS, a training-free approach to accelerate the inference of flow-matching-based TTS models. By inspecting the sampling trajectory of F5-TTS, we identify redundant steps and propose Empirically Pruned Step Sampling (EPSS), a non-uniform time-step sampling strategy that effectively reduces the number of sampling steps. Our approach achieves a 7-step generation with an inference RTF of 0.030 on an NVIDIA RTX 3090 GPU, making it 4 times faster than the original F5-TTS while maintaining comparable performance. Furthermore, EPSS performs well on E2 TTS models, demonstrating its strong generalization ability.
format Preprint
id arxiv_https___arxiv_org_abs_2505_19931
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Accelerating Flow-Matching-Based Text-to-Speech via Empirically Pruned Step Sampling
Zheng, Qixi
Chen, Yushen
Niu, Zhikang
Ma, Ziyang
Wang, Xiaofei
Yu, Kai
Chen, Xie
Audio and Speech Processing
Sound
Flow-matching-based text-to-speech (TTS) models, such as Voicebox, E2 TTS, and F5-TTS, have attracted significant attention in recent years. These models require multiple sampling steps to reconstruct speech from noise, making inference speed a key challenge. Reducing the number of sampling steps can greatly improve inference efficiency. To this end, we introduce Fast F5-TTS, a training-free approach to accelerate the inference of flow-matching-based TTS models. By inspecting the sampling trajectory of F5-TTS, we identify redundant steps and propose Empirically Pruned Step Sampling (EPSS), a non-uniform time-step sampling strategy that effectively reduces the number of sampling steps. Our approach achieves a 7-step generation with an inference RTF of 0.030 on an NVIDIA RTX 3090 GPU, making it 4 times faster than the original F5-TTS while maintaining comparable performance. Furthermore, EPSS performs well on E2 TTS models, demonstrating its strong generalization ability.
title Accelerating Flow-Matching-Based Text-to-Speech via Empirically Pruned Step Sampling
topic Audio and Speech Processing
Sound
url https://arxiv.org/abs/2505.19931