Key facts
- Maya Research is developing AI models that aim to speak and respond like native speakers across languages and cultural contexts.
- The startup believes voice is the primary interface for the next four to five billion users of AI.
- Maya Research raised $1.9 million in a funding round led by South Park Commons.
- The company plans to use the funds to train larger conversational models and expand deployment infrastructure.
- Maya's voice model, Maya 1, ranks sixth globally on the Speech Arena benchmark for conversational speech systems.
Maya Research, a startup founded by New York University graduates Dheemanth Reddy and Bharath Kumar Kakumani, is developing conversational AI models designed to function as native speakers across various languages, dialects, and cultural contexts. The company believes that for the next wave of internet users, voice is not merely a feature but the primary mode of interaction with technology.
The startup aims to differentiate itself in a crowded voice AI market by focusing on the nuances of conversation, including hesitation, affirmation, uncertainty, and emotion, rather than just speech generation. Reddy stated that current models "know how to talk, but they don't know what to talk about."
This development follows Maya Research's successful $1.9 million funding round, led by South Park Commons. The capital infusion is earmarked for training larger conversational models, enhancing deployment infrastructure, and deepening the understanding of user interactions with voice-first AI systems in emerging markets.
The company operates both a model platform and a consumer application. Its models have reportedly garnered over 440,000 downloads on Hugging Face, while its consumer app has surpassed 3 million downloads across India, Southeast Asia, and the Middle East and North Africa region. Telugu-speaking users constitute Maya's largest market, followed by users in Uttar Pradesh and West Bengal. The company also noted significant usage among women discussing a range of topics.
Maya's emotive voice model, Maya 1, is currently ranked sixth globally among open-weight models on Speech Arena, a benchmark for conversational speech systems, placing it on par with OpenAI's GPT-Realtime-2. The company emphasizes collecting regional variations and dialects by employing individuals to record conversations in villages and towns across India, believing cultural familiarity is key to voice AI adoption in emerging markets.
Looking ahead, Maya Research plans to generate revenue by serving as a discovery and navigation layer, helping users access financial products, government schemes, healthcare services, and other digital offerings.