Kyutai · Voice Agents
Moshi (paper)
Speech-text model that generates speech-to-speech directly instead of the usual VAD to ASR to LLM to TTS pipeline, modelling model and user audio as two parallel streams so turn-taking is implicit, with an Inner Monologue step predicting time-aligned text before audio tokens. 160ms theoretical latency, ~200ms on an L4 GPU; 10.7k stars, MIT plus Apache-2.0.
Open the source ↗Similar Audio & Speech releases
- Dessa
Block (Dessa research lab) · Speech-only conversation model · Week 45, 2026
- Eleven v4 and v4 Turbo
ElevenLabs · Expressive TTS Model · Week 41, 2026
- Gemini 3.8 Live
Google · Live Voice Agent · Week 39, 2026
- VoiceMem
Tsinghua University · Voice Memory · Week 35, 2026
- Gemini 3.5 Transcribe
Google · Speech Recognition · Week 35, 2026
- Gemini Mac Voice
Google · Voice Agents · Week 31, 2026
More from Kyutai
- Unmute
Kyutai · Voice Agents · Week 30, 2026
- Moshi Open-Source Release
Kyutai · Voice Agents · Week 30, 2026
- Kyutai
Kyutai · Music Generation · Week 30, 2026
- Hibiki-Zero Blog
Kyutai · Speech-to-Text · Week 7, 2026
- Hibiki-Zero
Kyutai · Speech-to-Text · Week 7, 2026
Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.