openclaw-voice

Transcribe audio to text and generate spoken AI responses using Whisper and ElevenLabs via CLI with transcript storage and search.

OpenClaw Voice Skill

Overview

Voice conversation with AI via Whisper STT + ElevenLabs TTS. Records audio, transcribes, generates spoken responses, stores transcripts. For v1: CLI-based commands only (no real-time conversation mode — that's v1.1). Focus on transcript storage/search + TTS/STT wrappers.

Tech Stack

• Node.js ESM
• better-sqlite3 (WAL mode always)
• commander for CLI
• uuid for IDs
• @openclaw/interchange (import from ../interchange/src/index.js)
• NO external audio packages — use child_process to call sox/rec and ffplay

Database

See migrations/001_initial.sql for schema.

CLI

See src/cli.js for commands.

Tests

Run with npm test.