JVS
JVS supports multi-speaker Japanese synthesis with recordings in normal, whispered and falsetto styles.
Visit official source (opens in a new tab)Research use
Use the common-text subsets for speaker comparisons and the style subsets for voice-conversion experiments.
Limitations
Voice styles differ in coverage and quantity. Check the publisher’s separate conditions for recordings and labels.
License and access
- Dataset license
- JVS corpus terms (opens in a new tab)
- Access status
- Direct access
The publisher links to a public Google Drive ZIP. Drive may ask you to confirm the large-file download. Review the separate terms for audio and labels.