This 30M-parameter hybrid CTC-RNNT Branchformer model, developed by BharatGen, is designed for Automatic Speech Recognition (ASR) in Hindi. Trained on over 1,000 hours of Hindi audio-text parallel data, the model achieves a Word Error Rate (WER) of 14.85% on the Vistaar benchmark, averaged across all included datasets.
For any queries, please visit https://bharatgen.discourse.group/invites/BcouFsKk4g
MIT
BharatGen
Automatic Speech Recognition Model
PyTorch
Restricted
Sector Agnostic
29/04/25 06:29:32
N.A
94.49 MB
MIT
© 2026 - Copyright AIKosh. All rights reserved. This portal is developed by National e-Governance Division for AIKosh mission.