The system's architecture allows it to whisper, modulate tone, and even sing -- creating an experience that feels more human than machine, Agarwal said. (Photo | Screengrab, X)
Tech

Jaipur's IIT student builds world's first speech-to-speech AI Model that can sing, whisper, feel

Luna AI will directly process audio to generate human-like speech instead of converting it to text and back, resulting in faster, more expressive, and emotionally aware conversations.

PTI

JAIPUR: Jaipur-based 25-year-old founder Sparsh Agrawal has unveiled one of the first speech-to-speech foundational AI models that can sing, whisper, pause, and respond with emotional intelligence -- all developed without big-tech infrastructure or venture capital funding.

Launched under his startup Pixa AI, Luna AI directly processes audio to generate human-like speech instead of converting it to text and back, resulting in faster, more expressive, and emotionally aware conversations.

The system's architecture allows it to whisper, modulate tone, and even sing -- creating an experience that feels more human than machine, Agarwal said.

He recently met with Union IT minister Ashwini Vaishnaw and got appreciation from industry leaders for his achievement.

"Where is India's AI? Every WhatsApp group, every conference hallway, every founder call asks the same question. Today, we're sharing the answer. Meet Luna, world's first speech-to-speech foundational AI model to unify audio, music and speech," Agarwal posted on X after launching the model.

Benchmark results show Luna outperforming leading global systems such as OpenAI's GPT-4 TTS and ElevenLabs, with 50 per cent lower latency and greater naturalness in speech output.

"I didn't have a research lab or a USD 100 million runway," Agrawal said.

"I borrowed GPUs, cloud credits, and even took on credit card debt to build Luna. This is proof that world-class technology can come from India – built with resourcefulness, not just resources."

Agarwal is an IIT-BHU graduate. His other teammates are Nitish Kartik, Apoorv Singh and Pratyush Kumar.

Backed by investors including Kunal Shah, Kunal Kapoor, and Nikhil Kamath, Pixa AI aims to make Luna the "voice layer" for global entertainment, wellness, and automotive ecosystems.

Early demand has already emerged from automakers, gaming platforms, and consumer AI companies.

"While most voice models are built for customer support, Luna is built for emotion," Agrawal said.

"It's not just about responding -- it's about feeling."

HCL co-founder Padma Bhushan awardee Ajai Chowdhry posted on X on this achievement.

"Sparsh has created an amazing product. India's first!," Chowdhry said.

Sudarshan Kamath, a known figure in voice AI research and founder of smallest.ai, said, "Truly amazing deep tech founders building silently from India!"

"We need to back more people like Sparsh who understand how AI works and are willing to think from first principles," he said.

Agrawal, the only solo founder selected from over 15,000 applicants to the WTFund, said Luna's creation is part of his mission to make India a hub for emotionally intelligent AI innovation.

‘Say no and we’ll remember’: Trump issues Greenland ultimatum to NATO at Davos, rejects use of force

India yet to take call on joining Trump's 'Board of Peace' for Gaza, say sources

After NMC action, 50 Vaishno Devi College MBBS students in limbo as BOPEE can’t accommodate them

Military power the ultimate arbiter, but will to use it is more important, says IAF Chief AP Singh

Lucknow woman plots to implicate husband in false cow slaughter case

SCROLL FOR NEXT