announcer

Announcer

Play TTS announcements through AirPlay speakers via Airfoil and ElevenLabs.

How It Works

1. Generate speech via ElevenLabs (high-quality opus → stereo MP3)
2. Connect to AirPlay speakers via Airfoil
3. Play an optional chime (gong) followed by the announcement
4. Disconnect speakers after playback

Requirements

• Airfoil (Rogue Amoeba) — running on the host Mac
• ElevenLabs API key — set ELEVENLABS_API_KEY env var
• ffmpeg — for audio format conversion
• elevenlabs skill — sibling skill for TTS generation

Configuration

User config lives at ~/clawd/announcer/config.json:

{
  "speakers": ["Living (2)", "Kitchen", "Office"],
  "excluded": ["Computer"],
  "elevenlabs": {
    "voice_id": "your-voice-id",
    "format": "opus_48000_192"
  },
  "audio": {
    "output_format": "mp3",
    "stereo": true,
    "sample_rate": 48000,
    "bitrate": "256k",
    "chime_file": "gong_stereo.mp3"
  },
  "airfoil": {
    "source": "System-Wide Audio",
    "connection_timeout_seconds": 30,
    "volume": 0.7
  }
}

Config Fields

Field	Description
speakers	AirPlay speaker names to connect
excluded	Speaker names to never connect
elevenlabs.voice_id	ElevenLabs voice to use
audio.chime_file	Chime sound file in assets/ (set null to disable)
airfoil.connection_timeout_seconds	Time to wait for speakers to connect
airfoil.volume	Speaker volume (0.0–1.0)

Usage

# Announce to all configured speakers
python3 skills/announcer/scripts/announce.py "Dinner is ready!"

# Announce to specific speakers only
python3 skills/announcer/scripts/announce.py "Wake up!" --speakers "Kids Room"

# Skip the chime
python3 skills/announcer/scripts/announce.py "Quick note" --no-gong

File Structure

announcer/
├── SKILL.md
├── assets/
│   └── gong_stereo.mp3      # Announcement chime
└── scripts/
    └── announce.py           # Main announcement script

User config (not part of skill):

~/clawd/announcer/
└── config.json               # Speaker list, voice, audio settings