JVS corpus: free Japanese multi-speaker voice corpus

• DOI: 10.48550/arxiv.1908.06248
• Publication Date: 2019-01-01
• Authors: Shinnosuke Takamichi; Kentaro Mitsui; Yuki Saito; Tomoki Koriyama; Naoko Tanji; Hiroshi Saruwatari
• Keywords: Corpus Linguistics

Thanks to improvements in machine learning techniques, including deep learning, speech synthesis is becoming a task. To accelerate research, we are developing Japanese voice corpora reasonably accessible from not only academic institutions but also commercial companies. In 2017, released the JSUT corpus, which contains 10 hours of reading-style uttered by single speaker, for end-to-end text-to-speech synthesis. For more general use e.g., conversion and multi-speaker modeling, this paper, construct JVS data 100 speakers three styles (normal, whisper, falsetto). The corpus 30 22 parallel normal voices. This paper describes how designed summarizes specifications. available at our project page.