Salvato in:
Dettagli Bibliografici
Autori principali: Li, Bohan, Wang, Hankun, Zhang, Situo, Guo, Yiwei, Yu, Kai
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2410.21951
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916605476208640
author Li, Bohan
Wang, Hankun
Zhang, Situo
Guo, Yiwei
Yu, Kai
author_facet Li, Bohan
Wang, Hankun
Zhang, Situo
Guo, Yiwei
Yu, Kai
contents The auto-regressive architecture, like GPTs, is widely used in modern Text-to-Speech (TTS) systems. However, it incurs substantial inference time, particularly due to the challenges in the next-token prediction posed by lengthy sequences of speech tokens. In this work, we introduce VADUSA, one of the first approaches to accelerate auto-regressive TTS through speculative decoding. Our results show that VADUSA not only significantly improves inference speed but also enhances performance by incorporating draft heads to predict future speech content auto-regressively. Furthermore, the inclusion of a tolerance mechanism during sampling accelerates inference without compromising quality. Our approach demonstrates strong generalization across large datasets and various types of speech tokens.
format Preprint
id arxiv_https___arxiv_org_abs_2410_21951
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Fast and High-Quality Auto-Regressive Speech Synthesis via Speculative Decoding
Li, Bohan
Wang, Hankun
Zhang, Situo
Guo, Yiwei
Yu, Kai
Audio and Speech Processing
Artificial Intelligence
Sound
68T07
The auto-regressive architecture, like GPTs, is widely used in modern Text-to-Speech (TTS) systems. However, it incurs substantial inference time, particularly due to the challenges in the next-token prediction posed by lengthy sequences of speech tokens. In this work, we introduce VADUSA, one of the first approaches to accelerate auto-regressive TTS through speculative decoding. Our results show that VADUSA not only significantly improves inference speed but also enhances performance by incorporating draft heads to predict future speech content auto-regressively. Furthermore, the inclusion of a tolerance mechanism during sampling accelerates inference without compromising quality. Our approach demonstrates strong generalization across large datasets and various types of speech tokens.
title Fast and High-Quality Auto-Regressive Speech Synthesis via Speculative Decoding
topic Audio and Speech Processing
Artificial Intelligence
Sound
68T07
url https://arxiv.org/abs/2410.21951