kyutai-labs/unmute

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1506 stars · 249 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Unmute is a system that lets any text LLM listen and speak by wrapping it with Kyutai's low-latency speech-to-text and text-to-speech models. It orchestrates a browser frontend, a websocket backend, STT/TTS servers, and an LLM (e.g. via OpenRouter or vLLM) into a real-time voice conversation pipeline.

Use cases

• build a real-time voice chatbot powered by any text LLM
• add speech input and spoken responses to a text-only LLM
• self-host a low-latency voice assistant with STT and TTS
• create a web app where users talk to an AI and hear it reply
• run a voice agent pipeline with vLLM or OpenRouter as the LLM backend

When to choose

• you want a full self-hostable voice conversation stack with low latency
• you already have or can pick any text LLM and want to add speech I/O
• you have a CUDA GPU with 16+ GB VRAM and run Linux or WSL

When to avoid

• you need macOS or native Windows support
• you have no GPU available
• you only need STT or TTS alone rather than a full voice loop
• you need aarch64/ARM deployment

Facets

application · maturity active

speech-recognition tts llm-inference websocket http-server speech-processing large-language-models artificial-intelligence python voice-assistant real-time-speech stt tts conversational-ai low-latency voice linux docker gpu web-server

Member repositories

For agents

markdown · JSON · MCP: product_card(name="kyutai-labs/unmute")