LibriTTS
A multi-speaker English TTS corpus derived from LibriSpeech with 585 hours of 24 kHz high-quality recordings from 2,456 speakers
Duration
585 hours
Languages
1
Sample Rate
24 kHz
Published
2019-04
Description
1Rebuilt from LibriSpeech's original sources (LibriVox MP3s and Project Gutenberg texts)
22,456 speakers, 585 hours of speech
3Improvements over LibriSpeech: 24 kHz sampling rate, sentence-level segmentation, includes both original and normalized text, excludes high background noise utterances
4Designed specifically for TTS research
Language Details
| Language | Duration |
|---|---|
| English | 585 hours |
Publisher