Indic language transcription backed by data

Get world-class speed and accuracy for major Indian languages, so your users experience natural, reliable transcription in the languages they actually speak.

Hindi

Hindi ASR built for 350M+ native speakers

Trained on 430.6 hours of real-world audio from OpenSLR, Gramvaani, Shrutilipi, Kathbath, Vaani, and proprietary data in just 35.9 hours on dual A100 GPUs.

Telugu

Telugu ASR built for 80M+ native speakers

Trained on 110.6 hours of real-world audio from Kathbath and Google Fleurs in just 13.43 hours on dual A100 GPUs.

Kannada

Kannada ASR built for 40M+ native speakers

Trained on 90.6 hours of real-world audio from OpenSLR, Kathbath, and Vaani in just 9.9 hours on dual A100 GPUs.

Bengali

Bengali ASR built for 230M+ native speakers

Trained on 110.6 hours of real-world audio from OpenSLR, Kathbath, Vaani, and Shrutilipi in just 12.9 hours on dual A100 GPUs.