English

ContextSpeech: Expressive and Efficient Text-to-Speech for Paragraph Reading

Computation and Language 2023-10-10 v2 Artificial Intelligence Audio and Speech Processing

Abstract

While state-of-the-art Text-to-Speech systems can generate natural speech of very high quality at sentence level, they still meet great challenges in speech generation for paragraph / long-form reading. Such deficiencies are due to i) ignorance of cross-sentence contextual information, and ii) high computation and memory cost for long-form synthesis. To address these issues, this work develops a lightweight yet effective TTS system, ContextSpeech. Specifically, we first design a memory-cached recurrence mechanism to incorporate global text and speech context into sentence encoding. Then we construct hierarchically-structured textual semantics to broaden the scope for global context enhancement. Additionally, we integrate linearized self-attention to improve model efficiency. Experiments show that ContextSpeech significantly improves the voice quality and prosody expressiveness in paragraph reading with competitive model efficiency. Audio samples are available at: https://contextspeech.github.io/demo/

Keywords

Cite

@article{arxiv.2307.00782,
  title  = {ContextSpeech: Expressive and Efficient Text-to-Speech for Paragraph Reading},
  author = {Yujia Xiao and Shaofei Zhang and Xi Wang and Xu Tan and Lei He and Sheng Zhao and Frank K. Soong and Tan Lee},
  journal= {arXiv preprint arXiv:2307.00782},
  year   = {2023}
}

Comments

5 pages, 4 figures, Proceedings of Interspeech 2023

R2 v1 2026-06-28T11:20:24.874Z