Kyutai · 2024– · full-duplex
Kyutai Moshi
A full-duplex spoken dialogue model: it listens and speaks on overlapping streams. That is a different object from ‘read this paragraph.’ The live demo is at moshi.chat; no static clip is on this bench.
Known for
Overlapping conversation, not turn-taking TTS.
Sample unavailable. No redistributable clip was on file. Listen on the official site rather than a fake waveform.
- Approach
- Spoken dialogue
- Year / era
- 2024
- License class
- open
- Cloning
- No
- Streaming
- Yes
- Emotion controls
- Yes
- Languages (tags)
- en · fr
- Official
- Source