Kyutai

What is Kyutai?

Kyutai is an artificial intelligence research lab headquartered in Paris whose stated goal is to build and democratize artificial general intelligence through open science. It releases models, code and demos.

Key capabilities of Kyutai

• Moshi: A speech-native dialogue system that processes speech directly
• Unmute: Lets any LLM listen and speak with streaming speech-to-text and text-to-speech
• Pocket TTS: A 100 million parameter text-to-speech model that runs on CPU faster than real time
• Hibiki: Speech-to-speech translation
• Helium 1: A 2B-parameter multilingual language model
• MIRA: A real-time world model

How Kyutai works

Developers download models and code from GitHub and Hugging Face and run them locally or on their own infrastructure. Moshi handles real-time conversation without converting speech to text first.

Who uses Kyutai?

Researchers and developers who want open speech models. Kyutai is funded by Iliad Group, CMA CGM Group and Schmidt Sciences.

Kyutai pricing

Kyutai releases models and code under open-source licenses at no cost. Check each repository for license terms.

Kyutai alternatives

Alternatives include Speechmatics and Gladia for speech APIs and Hume AI for expressive voice. Kyutai is notable for open weights and research releases.