Hugging Face releases fully local speech-to-speech stack for Reachy Mini
hugging face released a local speech-to-speech pipeline for reachy mini using llama.cpp, silero vad, and qwen3-tts to ensure privacy and eliminate api costs.
hugging face published a guide for deploying a fully local conversational backend for the reachy mini robot, utilizing a 'speech-to-speech' library that acts as a cascaded pipeline. the recommended configuration leverages llama.cpp to serve the gemma 4 model with a 64k context window, alongside silero for voice activity detection, parakeet-tdt for speech-to-text, and qwen3 for text-to-speech. this architecture exposes a websocket api compatible with the openai realtime protocol, enabling developers to swap components like the llm or tts models based on specific latency or quality needs while keeping all audio processing on local hardware.