IIT-M’s Bodhan AI Unveils Four Open AI Models for Indian Languages

Bodhan AI, an IIT Madras-incubated Centre of Excellence in AI for Education, has announced a suite of four open foundational AI models for Indian languages developed in collaboration with NVIDIA, according to a press release.

The initiative, undertaken alongside AI4Bharat, covers four core capabilities speech recognition, text-to-speech, machine translation and optical character recognition.The models include Indic-Transcribe for speech recognition, Indic-Speak for text-to-speech, Indic-Translate for machine translation and Indic-OCR for optical character recognition.

The models have been developed using the NVIDIA NeMo framework, which includes post-training of NVIDIA Nemotron 3.5 ASR to support Indian regional dialects and accents, according to the release.Inference for the models is served through NVIDIA TensorRT-LLM and vLLM microservices.

Bodhan AI and AI4Bharat said the initiative is focused on accelerating the adoption of multilingual AI solutions across the Indian language AI ecosystem, particularly in the education sector.

Prof Mitesh Khapra, Principal Investigator at Bodhan AI and AI4Bharat, said the partnership with NVIDIA would help bring open, state-of-the-art capabilities to developers.

The foundational models are being made available through open-weight releases and hosted APIs within India's sovereign digital ecosystem.According to the release, educational applications using these models will remain free for learners, teachers and partnering state governments.

The four models provide different language-processing capabilities. Indic-Transcribe is designed for speech recognition, Indic-Speak for converting text to speech, Indic-Translate for machine translation and Indic-OCR for recognising text from images and documents.

The initiative is being developed with a focus on Indian languages and education, with the models made accessible through open-weight releases and hosted application programming interfaces.

The announcement comes as AI applications in Indian languages continue to be developed for use across different sectors. Bodhan AI's initiative focuses specifically on education and multilingual language technologies.

The collaboration brings together Bodhan AI, AI4Bharat and NVIDIA for the development and deployment of the four foundational models.The models are available through the open-weight releases and hosted APIs within India's sovereign digital ecosystem, according to the release.