Fact · Languages and accents
7348hours of speech
The IndicVoices dataset contains 7,348 hours of read, extempore and conversational speech from 16,237 speakers, covering 22 Indian languages.
Checked against its source on October 17, 2026
In the source's own words
“a total of 7348 hours of read (9%), extempore (74%) and conversational (17%) audio from 16237 speakers covering 145 Indian districts and 22 languages”
IndicVoices: Towards building an Inclusive Multilingual Speech Dataset for Indian Languages, Tahir Javed et al., arXiv preprint (AI4Bharat, IIT Madras), 2024, Abstract.
Open the original- What it measures
- Most of the audio is extempore (unscripted) and conversational speech, which is closer to dictation than read text. Most of it is recorded from everyday speakers.
- What it does not show
- The dataset is for training and testing. Hours of audio do not tell you how accurate any consumer app is on a given language.