NEW DELHI: Bengaluru-based Gnani.ai today unveiled Inya VoiceOS, a 5-billion-parameter voice-to-voice foundational model built entirely in India to enable natural spoken interaction across multiple Indian languages. Prime Minister Narendra Modi inaugurated the system at the India AI Impact Summit 2026 at Bharat Mandapam, where it was released as a research preview under the government-backed IndiaAI Mission.
Unlike conventional speech systems that transcribe audio to text and back again, Inya VoiceOS processes audio directly in acoustic and semantic space, preserving tone, emotion, and pacing in real time. The architecture handles interruptions and overlapping speech without restarting conversations, a significant departure from pipeline-based systems that lose conversational context at each processing step.
The company targets government helplines, emergency response systems, and enterprise applications in banking, healthcare, and logistics. Gnani.ai currently processes around 10 million calls per day and counts HDFC Bank, Bank of Baroda, and IDFC First Bank among its customers.
The research preview precedes a larger 14-billion-parameter model the company plans to deploy across all 22 scheduled Indian languages within 18 months. Gnani.ai was selected under the IndiaAI Mission after the IT Ministry reviewed over 500 proposals, tasking the startup with building sovereign voice AI infrastructure built, trained, and deployed entirely within India.
Previous
