Technology
Archived — This article has been archived. The information may be outdated.
Bodhan AI releases open Indic language AI models
The Hindu BusinessLine··12 Sept
Bodhan AI, incubated at IIT Madras, announced open-weight foundational models for Indic languages built with AI4Bharat on Friday. The suite covers speech recognition, text-to-speech, machine translation and optical character recognition. Teams train using NVIDIA's NeMo framework for automatic speech recognition, translation and OCR, the Centre of Excellence in AI for Education said in its release.
Prism
What It Means For You
- Developers can fine-tune open-weight Indic models for local language apps without starting from scratch.
- Multilingual OCR and speech tools can lower costs for Indian language publishers.
- Education startups may adopt the models for classroom transcription and translation.
What's Happening
- Bodhan AI is a Centre of Excellence in AI for Education at IIT Madras.
- AI4Bharat contributed Indic datasets and multilingual modelling expertise.
- Principal investigator Mitesh Khapra said NVIDIA partnership speeds deployment.
Indic open models
- Open-weight releases let builders adapt models under their own infrastructure.
- Indic-Transcribe, Indic-Speak, Indic-Translate and Indic-OCR are the four capabilities.
- NeMo is NVIDIA's framework for speech and translation training.
all-news
