Gnani.AI to develop India's first multilingual, emotion-aware voice model
What's the story
Bengaluru-based tech start-up, Gnani.AI, was recently chosen under the IndiaAI Mission to build a sovereign foundational large language model (LLM).
The company is now working on an innovative voice-first, multilingual, and emotionally intelligent artificial intelligence (AI) system.
Ganesh Gopalan, co-founder and CEO of Gnani.AI, told Moneycontrol that they are developing India's first real-time voice-to-voice foundational model.
Technical advancement
'Latency kills voice AI adoption'
Gopalan emphasized the importance of reducing latency in voice AI adoption. He said, "Latency kills voice AI adoption."
To tackle this challenge, Gnani.AI is developing a "fused" architecture that combines speech recognition and language understanding into one step.
This new approach will result in faster, smoother, and more human-like AI conversations.
It is especially useful for applications where emotional nuances are critical, such as customer support or healthcare.
Data collection
Extensive data collection for nearly a decade
Gnani.AI has been collecting and curating millions of hours of audio across demographics for nearly a decade.
The company has a proprietary dataset that includes voice samples from every 780-odd districts in India, covering a wide range of Indic accents, dialects, and real-world noise environments.
This extensive data collection process has enabled Gnani.AI to create an AI model that not only understands languages but also how they are spoken.
Project involvement
Gnani.AI selected alongside Soket and Gan.AI
Gnani.AI was selected for the second cohort of AI start-ups under the IndiaAI Mission, along with Soket and Gan.AI.
The project was announced by IT Minister Ashwini Vaishnaw on January 30.
On February 11, the government allocated ₹1,500 crore under the mission to fund indigenous foundational models.
Details on compute access and rollout are still being finalized for start-ups selected under this phase.
Expansion
Use in rural UP to provide health guidance
Gnani.AI plans to use its foundational models in high-impact sectors such as education, healthcare, and agriculture.
One of these projects is already being piloted in rural Uttar Pradesh, where voice AI is being used to provide timely and emotionally supportive health guidance to expecting mothers.
The company hopes this project will help reduce India's high infant mortality rates.
Information
Models being built to solve India-specific problems
Gopalan highlighted that Gnani.AI is building these models to solve India-specific problems. He said, "A global model might work for English, and Hindi to some extent, but typically fails with other Indian languages."