Control the VTuber avatar system — speak through it with lip sync, change expressions, manage the avatar renderer and control server. Use when interacting with the avatar, making it speak, changing expressions, or troubleshooting avatar connection issues.
Sign in to like and favorite skills
~/openclaw/scripts/avatar-speak.sh "text" [emotion] [output]
Output controls where audio plays:
Default output is speakers.
neutral (default, eyes open), happy, sad, angry, relaxed, surprised
Use neutral by default. happy closes the eyes (anime smile) — only use for genuine excitement.
When avatar system is started, these are always available:
The bot chooses per-speak where audio goes. Virtual mic and camera are always-on pipes.
systemctl --user {start|stop|status|restart} avatar-control-server Renderer: cd ~/openclaw/avatar/renderer && npm run dev
Port 8765 — must send identify first: { type: "identify", role: "agent", name: "@agentName@" }
Commands after identify:
Wait ~1s after identify before sending commands. Wait for speakAck duration + 2s buffer before closing WebSocket.
See server.js in ~/openclaw/avatar/control-server/ for full protocol.
Agent sends speak with output target -> control server runs edge-tts -> generates MP3 -> ffmpeg plays to chosen PulseAudio sink(s) -> renderer gets lip sync data only (visual animation, no browser audio).