Voice Loop

Hands-free voice conversation: speak → Whisper transcribes (local) → OpenClaw streams response (SSE) → Kokoro speaks sentence-by-sentence (local).

Architecture

Microphone → Whisper STT (local, ~2s) → OpenClaw API (cloud, ~4-10s) → Kokoro TTS (local, <1s) → Speakers

Streaming TTS speaks each sentence as it arrives — first audio in ~3s, not 13s.

Setup

Run the setup script to install dependencies and download models:

bash scripts/setup.sh