SimpleSpeech: Towards Simple and Efficient Text-to-Speech with Scalar Latent Transformer Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Dongchao, Wang, Dingdong, Guo, Haohan, Chen, Xueyuan, Wu, Xixin, Meng, Helen
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914834185977856
author Yang, Dongchao
Wang, Dingdong
Guo, Haohan
Chen, Xueyuan
Wu, Xixin
Meng, Helen
author_facet Yang, Dongchao
Wang, Dingdong
Guo, Haohan
Chen, Xueyuan
Wu, Xixin
Meng, Helen
contents In this study, we propose a simple and efficient Non-Autoregressive (NAR) text-to-speech (TTS) system based on diffusion, named SimpleSpeech. Its simpleness shows in three aspects: (1) It can be trained on the speech-only dataset, without any alignment information; (2) It directly takes plain text as input and generates speech through an NAR way; (3) It tries to model speech in a finite and compact latent space, which alleviates the modeling difficulty of diffusion. More specifically, we propose a novel speech codec model (SQ-Codec) with scalar quantization, SQ-Codec effectively maps the complex speech signal into a finite and compact latent space, named scalar latent space. Benefits from SQ-Codec, we apply a novel transformer diffusion model in the scalar latent space of SQ-Codec. We train SimpleSpeech on 4k hours of a speech-only dataset, it shows natural prosody and voice cloning ability. Compared with previous large-scale TTS models, it presents significant speech quality and generation speed improvement. Demos are released.
format Preprint
id arxiv_https___arxiv_org_abs_2406_02328
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle SimpleSpeech: Towards Simple and Efficient Text-to-Speech with Scalar Latent Transformer Diffusion Models
Yang, Dongchao
Wang, Dingdong
Guo, Haohan
Chen, Xueyuan
Wu, Xixin
Meng, Helen
Sound
Audio and Speech Processing
In this study, we propose a simple and efficient Non-Autoregressive (NAR) text-to-speech (TTS) system based on diffusion, named SimpleSpeech. Its simpleness shows in three aspects: (1) It can be trained on the speech-only dataset, without any alignment information; (2) It directly takes plain text as input and generates speech through an NAR way; (3) It tries to model speech in a finite and compact latent space, which alleviates the modeling difficulty of diffusion. More specifically, we propose a novel speech codec model (SQ-Codec) with scalar quantization, SQ-Codec effectively maps the complex speech signal into a finite and compact latent space, named scalar latent space. Benefits from SQ-Codec, we apply a novel transformer diffusion model in the scalar latent space of SQ-Codec. We train SimpleSpeech on 4k hours of a speech-only dataset, it shows natural prosody and voice cloning ability. Compared with previous large-scale TTS models, it presents significant speech quality and generation speed improvement. Demos are released.
title SimpleSpeech: Towards Simple and Efficient Text-to-Speech with Scalar Latent Transformer Diffusion Models
topic Sound
Audio and Speech Processing
url https://arxiv.org/abs/2406.02328