speech-to-speech

Hugging Face's speech-to-speech is an open-source, modular voice-agent pipeline (VAD->STT->LLM->TTS) with an OpenAI Realtime-compatible WebSocket API, supporting swappable components and local or hosted models.