What it is
172k+ quality-checked Indian speech clips with verbatim native-script transcripts, grouped by primary language for multilingual ASR. 172k+ short 16 kHz mono speech clips across 12 language splits plus other, with verbatim transcripts, state, gender, and full spoken-language metadata from 12+ Indian states.
Key results
- 172k+ 16 kHz mono WAV clips
- 12 language splits plus an other bucket
- Fields include transcript, state, gender, languages; CC-BY-4.0
Useful for training and evaluating multilingual ASR on Indian regional languages and code-switched speech.