Gnani.ai has launched Vachana TTS, a foundational text to speech model with zero shot voice cloning across 12 Indian languages, under the IndiaAI Mission.
The company said the model delivers human quality speech synthesis with a Mean Opinion Score of 4.23 and a character error rate below 0.6 percent. Vachana TTS is the second model released under Inya VoiceOS and is positioned for national scale deployment. The first model launched is Vachana STT model that was launched yesterday.
The Vachana TTS model supports Hindi, Bengali, Tamil, Telugu, Kannada, Malayalam, Gujarati, Marathi, Punjabi, Odia, Assamese and Indian English. Gnani.ai said independent evaluations show Vachana TTS outperforms global text to speech providers on Indic language naturalness and pronunciation accuracy, while offering lower cost per character.
Vachana TTS allows zero shot voice cloning using less than 10 seconds of reference audio, preserving pitch, speech rate, timbre and speaking style. It also enables cross lingual voice cloning, allowing the same voice to speak across multiple Indian languages.
The system supports real time streaming for conversational applications and batch synthesis for large scale content generation, with predictable latency in high concurrency environments. Built and deployed entirely in India, the model complies with data localisation requirements.
The company said the model can be used for public announcements, emergency alerts, educational content, conversational AI agents and IVR systems. Vachana TTS is available immediately via API access, with 100000 free minutes for early adopters. On premises deployment is offered for regulated sectors.
Ganesh Gopalan, Co-Founder and CEO of Gnani.ai, said, “With Vachana TTS, we are bringing genuine emotion into synthesised speech - warmth, urgency, empathy - delivered with significantly better accuracy than anything built for Indian languages before, and at a price point that makes it accessible to every enterprise and government body in the country. When a citizen receives an emergency alert or a customer hears a support agent, the voice must feel human. That is the standard we have built to."
Inya VoiceOS is India’s first five-billion-parameter voice-to-voice foundational AI model. It will sit alongside the company’s existing stack, which already powers voice AI agents, speech models and vertical-specific small language models for banks, insurers, non-banking financial companies, retailers and consumer businesses across India.
Your reaction