SpeechTokenizer: Unified Speech Tokenizer for Speech Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Xin, Zhang, Dong, Li, Shimin, Zhou, Yaqian, Qiu, Xipeng
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910305573928960
author Zhang, Xin
Zhang, Dong
Li, Shimin
Zhou, Yaqian
Qiu, Xipeng
author_facet Zhang, Xin
Zhang, Dong
Li, Shimin
Zhou, Yaqian
Qiu, Xipeng
contents Current speech large language models build upon discrete speech representations, which can be categorized into semantic tokens and acoustic tokens. However, existing speech tokens are not specifically designed for speech language modeling. To assess the suitability of speech tokens for building speech language models, we established the first benchmark, SLMTokBench. Our results indicate that neither semantic nor acoustic tokens are ideal for this purpose. Therefore, we propose SpeechTokenizer, a unified speech tokenizer for speech large language models. SpeechTokenizer adopts the Encoder-Decoder architecture with residual vector quantization (RVQ). Unifying semantic and acoustic tokens, SpeechTokenizer disentangles different aspects of speech information hierarchically across different RVQ layers. Furthermore, We construct a Unified Speech Language Model (USLM) leveraging SpeechTokenizer. Experiments show that SpeechTokenizer performs comparably to EnCodec in speech reconstruction and demonstrates strong performance on the SLMTokBench benchmark. Also, USLM outperforms VALL-E in zero-shot Text-to-Speech tasks. Code and models are available at https://github.com/ZhangXInFD/SpeechTokenizer/.
format Preprint
id arxiv_https___arxiv_org_abs_2308_16692
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle SpeechTokenizer: Unified Speech Tokenizer for Speech Large Language Models
Zhang, Xin
Zhang, Dong
Li, Shimin
Zhou, Yaqian
Qiu, Xipeng
Computation and Language
Sound
Audio and Speech Processing
Current speech large language models build upon discrete speech representations, which can be categorized into semantic tokens and acoustic tokens. However, existing speech tokens are not specifically designed for speech language modeling. To assess the suitability of speech tokens for building speech language models, we established the first benchmark, SLMTokBench. Our results indicate that neither semantic nor acoustic tokens are ideal for this purpose. Therefore, we propose SpeechTokenizer, a unified speech tokenizer for speech large language models. SpeechTokenizer adopts the Encoder-Decoder architecture with residual vector quantization (RVQ). Unifying semantic and acoustic tokens, SpeechTokenizer disentangles different aspects of speech information hierarchically across different RVQ layers. Furthermore, We construct a Unified Speech Language Model (USLM) leveraging SpeechTokenizer. Experiments show that SpeechTokenizer performs comparably to EnCodec in speech reconstruction and demonstrates strong performance on the SLMTokBench benchmark. Also, USLM outperforms VALL-E in zero-shot Text-to-Speech tasks. Code and models are available at https://github.com/ZhangXInFD/SpeechTokenizer/.
title SpeechTokenizer: Unified Speech Tokenizer for Speech Large Language Models
topic Computation and Language
Sound
Audio and Speech Processing
url https://arxiv.org/abs/2308.16692