The People's Speech
A 30,000-hour large-scale diverse English ASR dataset under CC-BY-SA license, supporting both academic and commercial use
Duration
30000 hours
Languages
1
Sample Rate
16 kHz
Published
2021-11
Description
1Collected existing audio from the Internet Archive and force-aligned with text
2Diverse sources: movies, lectures, historical recordings, podcasts, etc.
3Contains real environmental noise and diverse accents
4CC-BY-SA license (with CC-BY subset), supports commercial use
5Data collection cost reduced from an estimated $5 million to approximately $3,000
Language Details
| Language | Duration |
|---|---|
| English | 30000 hours |
Publisher