Gnani.AI, one of many AI startups chosen as a part of the IndiaAI Mission, will launch its voice basis mannequin on the India AI Influence Summit subsequent week, its co-founder and chief expertise officer, Ananth Nagaraj, stated.
The mannequin is constructed on a 14-billion-parameter voice AI basis mannequin, delivering multilingual, real-time speech processing with superior reasoning capabilities. The mannequin is designed for low-latency, speech-to-speech communication and is meant for purposes in buyer help, training, accessibility and public-facing programs.
“We’re launching our voice-to-voice basis mannequin in six languages and goal to develop it to all 22 languages within the subsequent 18 months,” Nagaraj stated. The six languages are English, Hindi, Kannada, Telugu, Tamil and Gujarati.
Gnani will even launch a multilingual text-to-speech mannequin that has the flexibility to clone voices hyper-realistically. Named Vachana STT, it has been skilled on over a million hours of real-world voice knowledge, spanning greater than 1,056 domains.
“Will probably be a voice plus avatar. So should you take a look at our avatar additionally, the place we’ll use a customer support consultant, it’s a digital twin of her or him. So in our voice-to-voice mannequin or speech-to-text and the LLM understanding layer, together with text-to-speech, all of them will likely be driving this avatar.”
Voice AI is predicted to be the one sensible interface to herald true digital equality in India and the subsequent massive factor, in response to Infosys chairman Nandan Nilekani. “Simply as UPI made digital funds easy for everybody, voice-driven interfaces can take away boundaries to alternative in sectors reminiscent of agriculture, training and others for each citizen. Literacy will not be a barrier,” he stated final month.
When requested about this, Nagaraj agreed.
“Human-machine interplay goes to be by voice. And India being so various, there’s going to be voice AI in their very own mom tongue. I feel with voice AI, when it unlocks human-machine interplay by native languages, it would expose virtually 100 crore customers to the expertise.”
Sarvam AI can be hopeful of popping out with its sovereign AI mannequin on the similar time, its co-founder Vivek Raghavan stated in November.
Moreover Sarvam and Gnani, Soket will develop India’s first open-source 120-billion-parameter basis mannequin optimised for the nation’s linguistic variety, concentrating on sectors reminiscent of defence, healthcare and training. And Gan AI will create a 70-billion-parameter multilingual basis mannequin concentrating on text-to-speech capabilities.
Gnani, which counts Samsung Enterprise Funding amongst its traders, expects to finish this monetary yr with income of $20 million, up from simply ₹56 crore a yr earlier, as it really works with extra monetary companies corporations reminiscent of HDFC Financial institution, Financial institution of Baroda and IDFC First Financial institution.
