Archive · 30 Oct 2025
Pixa AI Launches Luna, India's First Speech-to-Speech AI Model
Imagine a computer that can talk like a real person, understanding not just your words but also how you feel!
Pixa AI, a company in India, has created something called Luna AI.
Instead of turning speech into text and then back into speech, Luna understands and speaks directly, making conversations much faster and more natural.
It can even change its voice to sound happy, sad, or whisper.
This AI is trained on millions of hours of voice data and is already better than some big AI systems at sounding natural.
Pixa AI is working with companies in car entertainment, toys, and customer service.
They also plan to make Luna speak many languages soon.
The goal is to make AI feel more like talking to a friend than a robot.
Pixa AI has launched Luna AI, India's first speech-to-speech foundational voice model.
Luna AI processes audio directly, enabling real-time, emotionally intelligent responses without text conversion, including singing and whispering.
Internal benchmarks show Luna AI outperforms leading systems like OpenAI, Deepgram, and ElevenLabs in accuracy and speech naturalness.
Pixa AI is targeting B2B applications in entertainment, automotive, gaming, and customer service, with plans for a conversational AI for Indians.
The startup has secured seed funding in the single-digit million range and plans to expand multilingual support soon.
- Who
- Pixa AI, an Indian startup founded by Sparsh Agrawal
- What
- Launched Luna AI, a speech-to-speech foundational voice model capable of real-time, emotionally intelligent speech generation, including singing and whispering.
- When
- Announced October 30, 2025
- Where
- India
- Why
- To move beyond robotic AI interactions and create conversational AI with artificial emotional intelligence, aiming to advance applications in various sectors.
Current AI Voice Systems
Pixa AI's Luna AI
Processing Method
Current AI Voice Systems
Typically convert speech to text, then text to speech, introducing latency.
Pixa AI's Luna AI
Processes speech directly to speech, eliminating conversion steps for faster, more natural interaction.
Emotional Intelligence
Current AI Voice Systems
Often perceived as robotic or lacking emotional nuance.
Pixa AI's Luna AI
Designed for emotional intelligence, capable of understanding and expressing emotions like happiness, sadness, and sarcasm.
Capabilities
Current AI Voice Systems
Primarily focused on verbal communication.
Pixa AI's Luna AI
Can sing, whisper, pause, and respond with contextual emotional nuance.
- Startup Name
- Pixa AI
- Product
- Luna AI (Speech-to-Speech Foundational Model)
- Founder
- Sparsh Agrawal
- Location
- India
- Core Technology
- Direct audio processing for real-time, emotionally intelligent speech generation
- Training Data
- Millions of hours of synthetic and open-source speech data
- Funding
- Single-digit million range
- Key Backers
- Nikhil Kamath, Kunal Shah, Kunal Kapoor
