• New AI model targets high-accuracy speech recognition across major Indian languages and accents
• Trained on over one million hours of real-world voice data for noisy, everyday use cases
• Supports India’s goal of building sovereign, homegrown AI infrastructure under the IndiaAI Mission

Bengaluru-based generative AI startup Gnani.ai has launched Vachana STT, a speech-to-text model designed specifically for Indian languages, under the government-backed IndiaAI Mission. The model is a core component of the company’s upcoming voice technology stack, VoiceOS.

Gnani.ai said Vachana STT has been trained on more than one million hours of real-world voice data, enabling it to deliver lower error rates compared to existing speech recognition systems. The startup claims the model performs especially well in noisy environments such as call centers and customer support operations, where audio quality is often inconsistent.

The model supports a wide range of Indian languages, including Hindi, Tamil, Telugu, Bengali, Kannada, Malayalam, Marathi, Gujarati, Punjabi, Odia, and Assamese. According to the company, Vachana STT has been trained across over 1,000 domains and does not require additional fine-tuning for different industry use cases.

Gnani.ai said the model is already being used by enterprises in sectors such as banking, telecom, and customer service. It currently processes around 10 million calls per day and offers near real-time transcription. The system is available to enterprises through APIs, with limited free access for early users.

Also Read: SquadStack.ai Launches In-App Voice AI Assistant to Reimagine Digital Customer Journeys

The launch is part of the IndiaAI Mission, under which a select group of startups has been chosen to build core AI models within the country. Gnani.ai was selected after the IT Ministry reviewed more than 500 proposals submitted earlier this year. The startup has been tasked with developing large-scale voice AI models focused on Indian languages and real-world speech scenarios.

As part of this mandate, Gnani.ai is also working on a 14-billion-parameter voice AI model for multilingual, real-time speech processing. The initiative aims to reduce reliance on foreign AI systems by building India-specific, homegrown AI infrastructure supported by national compute resources.

Read More News :

GABIT Buys Swedish Nutrition Brand Nack to Expand Into Supplements

Ather Energy Enters Auto Insurance Distribution