Kathbath
Contributors in India read text in 12 languages for a multilingual speech benchmark.
Visit official source (opens in a new tab)Research use
Compare low-resource adaptation across languages and evaluate speakers shared with or held out from training.
Limitations
The corpus uses prompted reading. AI4Bharat describes separate rights for the original IndicCorp text.
License and access
- Dataset license
- CC0 (publisher data, with separate source-text rights) (opens in a new tab)
- Access status
- Direct access
Use the audio and transcript archive links in the AI4Bharat repository’s download tables.