Miso One

Miso One is an open-weights, 8B-parameter text-to-speech (TTS) system from Miso Labs, built to generate highly realistic, expressive, and emotionally varied English conversational speech. Instead of flat, robotic output, it brings natural warmth and intonation to spoken language, making it a strong fit for voice-agent research and everyday developer workflows.

Under the hood, Miso One is built on a Sesame-style conversational speech model (CSM) architecture with Mimi audio codes, paired with a highly optimized inference capability that delivers a published low latency of just 110 ms. That level of speed makes it well suited to real-time, interactive applications where responsiveness truly matters.

Beyond standard text-to-speech generation, Miso One also supports voice continuation and one-shot voice cloning from audio context, with clear consent boundaries in place. This gives researchers and developers flexible tools for creating consistent, expressive voices while respecting ethical safeguards.