Meet Luna: IITian's AI model that sings, offers human-like interactions
What's the story
Sparsh Agrawal, a 25-year-old IIT-BHU graduate and entrepreneur from Jaipur, has launched Luna AI. The innovative model is one of the first speech-to-speech foundational artificial intelligence (AI) systems that can sing, whisper, pause, and respond with emotional intelligence. The groundbreaking project was developed by Agrawal's start-up Pixa AI without backing from big tech companies, though it did receive support from individual investors and venture capitalists.
Technological advancement
Quicker, more natural, and emotionally aware conversations
Luna AI is a major step forward in emotional intelligence, processing audio directly to create realistic speech. This method allows for quicker, more natural, and emotionally aware conversations. The model can even adjust its tone or sing, making interactions feel "more human than machine," according to Agrawal. He recently met Union IT Minister Ashwini Vaishnaw, who praised his innovation along with several industry leaders.
Performance comparison
Outperforming global systems like GPT-4 TTS and ElevenLabs
In early benchmark tests, Luna AI has outperformed global systems such as OpenAI's GPT-4 TTS and ElevenLabs by offering 50% lower latency and more natural-sounding speech. Agrawal stressed that he didn't have a research lab or a $100 million budget to build Luna. "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," he said.
Market response
Pixa AI plans to make Luna the 'voice layer' for...
Pixa AI, backed by investors like Kunal Shah, Kunal Kapoor, and Nikhil Kamath, plans to make Luna the "voice layer" for global entertainment, wellness, and automotive ecosystems. The model has already drawn interest 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."
Innovation
What's different about Luna AI?
Unlike other Indian voice models like Gnani and Sarvam that focus on automating customer-service interactions or text-to-speech systems, Luna AI is a pure speech-to-speech system. "We believe the future of voice AI is not automating call support centers but building artificial emotional intelligence," Agrawal said. The model focuses on expressivity, multiple accents, tonal variety, and an entertainment-first design instead of a monotonous robotic voice output.
Feature
Current pilots and potential applications of Luna
Luna AI is a pure voice-to-voice model that detects tonality in real-time if you're angry or calm, and adjusts its tone, pace, and delivery dynamically. The start-up is currently running proof-of-concept pilots with an Indian automobile manufacturer and a US-based wellness start-up where Luna powers interactive storytelling and AI-driven in-car entertainment. The company has also seen interest from customer-service sectors seeking emotionally adaptive sales or support voice agents.