Two 23-year-olds in Bengaluru have built Maya1, an AI text-to-speech model ranked 2nd among open-weight voice AIs
Unlike big tech models, Maya1 lets users create voices using natural prompts like “calm, elderly male teacher”
Maya1 adds over 20 emotions — laughter, sighs, anger — and speaks with human-like fluency under 100ms latency
From podcasts to gaming and accessibility tools, Maya1 brings expressive, low-lag voices to varied applications
Developed using free Cloud credits and rural speech data, Maya1 proves cutting-edge AI doesn’t need deep pockets
Maya Research is building a vast Indic speech dataset, aiming for AI that speaks in India’s languages and accents