Kyutai is a nonprofit AI lab releasing models, code, and demos that make real-time speech, multimodal interaction, and world simulation accessible through open science.
Products & media
Open speech-text model and dialogue framework supports natural, simultaneous listening and speaking for real-time voice conversations.
Streaming speech recognition models transcribe live conversations with low latency and batched processing for concurrent voice applications.
Streaming translation model converts French speech into English speech and text while the speaker continues talking.
Lightweight multilingual speech synthesis runs on CPUs, offering streaming output and voice cloning without a GPU service.
Streaming speech synthesis model produces low-latency audio for server-hosted voice applications and conversational systems.
Open-source voice interface connects speech recognition and synthesis to a text language model for real-time spoken interaction.
Cosigns
“After using it IRL, it feels magical to have such a quick response.”
