Enabling Beam Search for Language Model-Based Text-to-Speech Synthesis

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tu, Zehai, Zhang, Guangyan, Lu, Yiting, Adigwe, Adaeze, King, Simon, Guo, Yiwen
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910581885239296
author Tu, Zehai
Zhang, Guangyan
Lu, Yiting
Adigwe, Adaeze
King, Simon
Guo, Yiwen
author_facet Tu, Zehai
Zhang, Guangyan
Lu, Yiting
Adigwe, Adaeze
King, Simon
Guo, Yiwen
contents Tokenising continuous speech into sequences of discrete tokens and modelling them with language models (LMs) has led to significant success in text-to-speech (TTS) synthesis. Although these models can generate speech with high quality and naturalness, their synthesised samples can still suffer from artefacts, mispronunciation, word repeating, etc. In this paper, we argue these undesirable properties could partly be caused by the randomness of sampling-based strategies during the autoregressive decoding of LMs. Therefore, we look at maximisation-based decoding approaches and propose Temporal Repetition Aware Diverse Beam Search (TRAD-BS) to find the most probable sequences of the generated speech tokens. Experiments with two state-of-the-art LM-based TTS models demonstrate that our proposed maximisation-based decoding strategy generates speech with fewer mispronunciations and improved speaker consistency.
format Preprint
id arxiv_https___arxiv_org_abs_2408_16373
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Enabling Beam Search for Language Model-Based Text-to-Speech Synthesis
Tu, Zehai
Zhang, Guangyan
Lu, Yiting
Adigwe, Adaeze
King, Simon
Guo, Yiwen
Sound
Audio and Speech Processing
Tokenising continuous speech into sequences of discrete tokens and modelling them with language models (LMs) has led to significant success in text-to-speech (TTS) synthesis. Although these models can generate speech with high quality and naturalness, their synthesised samples can still suffer from artefacts, mispronunciation, word repeating, etc. In this paper, we argue these undesirable properties could partly be caused by the randomness of sampling-based strategies during the autoregressive decoding of LMs. Therefore, we look at maximisation-based decoding approaches and propose Temporal Repetition Aware Diverse Beam Search (TRAD-BS) to find the most probable sequences of the generated speech tokens. Experiments with two state-of-the-art LM-based TTS models demonstrate that our proposed maximisation-based decoding strategy generates speech with fewer mispronunciations and improved speaker consistency.
title Enabling Beam Search for Language Model-Based Text-to-Speech Synthesis
topic Sound
Audio and Speech Processing
url https://arxiv.org/abs/2408.16373