Paris-based nonprofit AI research lab focused on open-weights speech and language models. Best known for Moshi, a full-duplex speech-native dialogue model that processes audio directly without text conversion, and Pocket TTS, a 100M-parameter voice-cloning model that runs in real time on CPU with no GPU required.
Kyutai's CADRE replaces the laggy video autoencoders inherited from text-to-video models, letting interactive world models respond in real time on one GPU.
Kyutai trained its 100M-parameter on-device text-to-speech model with a new one-step generative objective, hitting sub-1% WER without Jacobian-vector products.
Kyutai and Mirelo release MuScriptor, an open-weight model that converts any multi-instrument recording into per-instrument MIDI using a novel three-stage training pipeline.
Kyutai's Pocket TTS adds French, Spanish, Portuguese, Italian, and German to its 100M-parameter model that runs faster than real-time on a laptop CPU with no GPU required.