Saved in:
Bibliographic Details
Main Authors: Luo, Dan, Ma, Chengyuan, Li, Weiqin, Wang, Jun, Chen, Wei, Wu, Zhiyong
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.10309
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913792691011584
author Luo, Dan
Ma, Chengyuan
Li, Weiqin
Wang, Jun
Chen, Wei
Wu, Zhiyong
author_facet Luo, Dan
Ma, Chengyuan
Li, Weiqin
Wang, Jun
Chen, Wei
Wu, Zhiyong
contents With the advancement of speech synthesis technology, users have higher expectations for the naturalness and expressiveness of synthesized speech. But previous research ignores the importance of prompt selection. This study proposes a text-to-speech (TTS) framework based on Retrieval-Augmented Generation (RAG) technology, which can dynamically adjust the speech style according to the text content to achieve more natural and vivid communication effects. We have constructed a speech style knowledge database containing high-quality speech samples in various contexts and developed a style matching scheme. This scheme uses embeddings, extracted by Llama, PER-LLM-Embedder,and Moka, to match with samples in the knowledge database, selecting the most appropriate speech style for synthesis. Furthermore, our empirical research validates the effectiveness of the proposed method. Our demo can be viewed at: https://thuhcsi.github.io/icme2025-AutoStyle-TTS
format Preprint
id arxiv_https___arxiv_org_abs_2504_10309
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AutoStyle-TTS: Retrieval-Augmented Generation based Automatic Style Matching Text-to-Speech Synthesis
Luo, Dan
Ma, Chengyuan
Li, Weiqin
Wang, Jun
Chen, Wei
Wu, Zhiyong
Sound
Artificial Intelligence
With the advancement of speech synthesis technology, users have higher expectations for the naturalness and expressiveness of synthesized speech. But previous research ignores the importance of prompt selection. This study proposes a text-to-speech (TTS) framework based on Retrieval-Augmented Generation (RAG) technology, which can dynamically adjust the speech style according to the text content to achieve more natural and vivid communication effects. We have constructed a speech style knowledge database containing high-quality speech samples in various contexts and developed a style matching scheme. This scheme uses embeddings, extracted by Llama, PER-LLM-Embedder,and Moka, to match with samples in the knowledge database, selecting the most appropriate speech style for synthesis. Furthermore, our empirical research validates the effectiveness of the proposed method. Our demo can be viewed at: https://thuhcsi.github.io/icme2025-AutoStyle-TTS
title AutoStyle-TTS: Retrieval-Augmented Generation based Automatic Style Matching Text-to-Speech Synthesis
topic Sound
Artificial Intelligence
url https://arxiv.org/abs/2504.10309