Saved in:
Bibliographic Details
Main Authors: Luo, Tianze, Miao, Xingchen, Duan, Wenbo
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.16689
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913749365948416
author Luo, Tianze
Miao, Xingchen
Duan, Wenbo
author_facet Luo, Tianze
Miao, Xingchen
Duan, Wenbo
contents Flow matching offers a robust and stable approach to training diffusion models. However, directly applying flow matching to neural vocoders can result in subpar audio quality. In this work, we present WaveFM, a reparameterized flow matching model for mel-spectrogram conditioned speech synthesis, designed to enhance both sample quality and generation speed for diffusion vocoders. Since mel-spectrograms represent the energy distribution of waveforms, WaveFM adopts a mel-conditioned prior distribution instead of a standard Gaussian prior to minimize unnecessary transportation costs during synthesis. Moreover, while most diffusion vocoders rely on a single loss function, we argue that incorporating auxiliary losses, including a refined multi-resolution STFT loss, can further improve audio quality. To speed up inference without degrading sample quality significantly, we introduce a tailored consistency distillation method for WaveFM. Experiment results demonstrate that our model achieves superior performance in both quality and efficiency compared to previous diffusion vocoders, while enabling waveform generation in a single inference step.
format Preprint
id arxiv_https___arxiv_org_abs_2503_16689
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle WaveFM: A High-Fidelity and Efficient Vocoder Based on Flow Matching
Luo, Tianze
Miao, Xingchen
Duan, Wenbo
Sound
Computation and Language
Audio and Speech Processing
Flow matching offers a robust and stable approach to training diffusion models. However, directly applying flow matching to neural vocoders can result in subpar audio quality. In this work, we present WaveFM, a reparameterized flow matching model for mel-spectrogram conditioned speech synthesis, designed to enhance both sample quality and generation speed for diffusion vocoders. Since mel-spectrograms represent the energy distribution of waveforms, WaveFM adopts a mel-conditioned prior distribution instead of a standard Gaussian prior to minimize unnecessary transportation costs during synthesis. Moreover, while most diffusion vocoders rely on a single loss function, we argue that incorporating auxiliary losses, including a refined multi-resolution STFT loss, can further improve audio quality. To speed up inference without degrading sample quality significantly, we introduce a tailored consistency distillation method for WaveFM. Experiment results demonstrate that our model achieves superior performance in both quality and efficiency compared to previous diffusion vocoders, while enabling waveform generation in a single inference step.
title WaveFM: A High-Fidelity and Efficient Vocoder Based on Flow Matching
topic Sound
Computation and Language
Audio and Speech Processing
url https://arxiv.org/abs/2503.16689