IISc SraVaani is pushing India’s voice AI capabilities beyond the 22 mainstream languages commonly supported by existing systems. Developed by the SPIRE Lab at the Indian Institute of Science (IISc), in collaboration with ARTPARK and with support from Google, the open speech-recognition model covers 65 Indian languages and dialects. SraVaani is built on more than 31,000 hours of natural speech collected through Project Vaani from people across 165 districts. The dataset captures regional accents, dialects and everyday speech patterns, helping to address the shortage of data for low-resource Indian languages.The model is available with demo and fine-tuning code, allowing researchers, startups and developers to adapt it for specific languages and applications. Potential uses include education, healthcare, government services and accessibility. The initiative highlights IISc’s role in developing inclusive AI technology for India’s diverse linguistic landscape.
IISc SraVaani: AI Model Expands Voice Tech to 65 Indian Languages
