Kyutai Moshi vs Voicebox

A side-by-side look at pricing, capabilities, pros, cons, and our editorial scores.

	Kyutai Moshi Audio	Voicebox Audio
Tagline	Open-source, full-duplex speech-to-speech foundation model with sub-200ms latency	Open-source desktop voice studio for local cloning, dictation, and giving MCP agents a voice.
Category	Audio	Audio
Pricing	Free· Free and open source. Models under CC-BY 4.0, code under MIT (Python) / Apache 2.0 (Rust). Self-hosted only — you pay your own compute (24GB+ GPU for PyTorch, or Apple Silicon via MLX).	Free· Free and open source; optional $VOICEBOX token donations
Model	Moshi (7B-class speech-text foundation model) + Mimi neural audio codec, in-house by Kyutai	Multi-model (Chatterbox, Qwen TTS, Whisper, etc.)
Editorial score	—	7.4 / 10
Use cases	Real-time voice assistant prototypesResearch on full-duplex spoken dialogueOn-device voice interaction on Apple Silicon via MLXLow-latency conversational agents behind WebSocketNeural audio codec experimentation with MimiSelf-hosted voice interface for privacy-sensitive appsSpeech tokenizer for downstream audio LLM trainingInterruptible in-car or wearable voice UX	voice-cloningtext-to-speechdictationagent-voicesmulti-voice-narration
Pros	Truly full-duplex — handles interruptions, overlap and back-channels rather than rigid turn-taking Sub-200ms practical latency on a single L4 GPU, well below third-party voice APIs Fully open weights (CC-BY 4.0) plus MIT/Apache code — self-host with no per-minute billing Ships with Mimi, a streaming neural audio codec that beats SpeechTokenizer and SemantiCodec Multiple inference backends: PyTorch for research, Rust/Candle for production, MLX for on-device Mac/iPhone Inner-monologue text prediction gives you a transcript alongside the audio stream for free	Fully local inference on Metal, CUDA, ROCm, Intel Arc, or DirectML Clones a voice from as little as 3 seconds of audio MCP server lets Claude Code, Cursor, Cline speak in cloned voices Bundles seven TTS engines, Whisper dictation, and a multi-track editor Open source with Mac, Windows, and Linux builds
Cons	English-only voices at launch — no multilingual support out of the box Knowledge and reasoning quality trail top text LLMs; it's a 7B-class model, not GPT-4o Voice Requires a 24GB+ GPU for the reference PyTorch build; on-device is only viable via MLX on Apple Silicon No hosted API or SaaS tier — you own the ops, scaling and safety filtering Only two fixed synthetic voices (Moshiko/Moshika); no voice cloning or speaker conditioning in the release	Desktop-only — no hosted/cloud option for non-GPU users Quality scales with local hardware; small models trade fidelity Shipped celebrity voice presets invite obvious consent concerns Young project (v0.2.0) with rough edges likely
Website	kyutai.org	voicebox.sh

Pick Kyutai Moshi if

✅ Truly full-duplex — handles interruptions, overlap and back-channels rather than rigid turn-taking
✅ Sub-200ms practical latency on a single L4 GPU, well below third-party voice APIs
✅ Fully open weights (CC-BY 4.0) plus MIT/Apache code — self-host with no per-minute billing
✅ Ships with Mimi, a streaming neural audio codec that beats SpeechTokenizer and SemantiCodec

Pick Voicebox if

✅ Fully local inference on Metal, CUDA, ROCm, Intel Arc, or DirectML
✅ Clones a voice from as little as 3 seconds of audio
✅ MCP server lets Claude Code, Cursor, Cline speak in cloned voices
✅ Bundles seven TTS engines, Whisper dictation, and a multi-track editor

Compare a different pair →