Kokoro TTS

Kokoro is an open-weight TTS model you can run locally or test in a hosted browser demo right now. The score here is unusually high for a model demo, which tracks — the voice output is noticeably more natural than what you get from most open models at this size class, and the gap with commercial APIs like ElevenLabs or OpenAI TTS has visibly narrowed. What matters for a builder: this is an MIT-licensed model, meaning you can ship it in a product without a per-character bill hitting you at scale. The demo lets you test multiple voice styles before you commit to setting anything up locally. Reservation: running it in production at low latency still requires real infrastructure work — this is not a drop-in API, it is a starting point. The gap with the best commercial voices is still audible on longer, expressive passages. -> Best for: AI engineer or solo founder who needs to cut TTS costs at scale