Human setup (one-time)

1. Create or edit ~/.openclaw/.env on the machine running OpenClaw.
2. Add your credentials (do not commit these):

ELEVENLABS_API_KEY=your_elevenlabs_key
# Optional (recommended): default voice
ELEVENLABS_VOICE_ID=your_voice_id

3. Restart the OpenClaw gateway (so the service environment picks up changes).

Inputs to collect (if missing)

• Text to speak.
• Optional: voiceId (ElevenLabs voice id). If not provided, use the default.
• Optional: output format (mp3 default).

Requirements (credentials)

This skill does not embed secrets. Credentials must be provided at runtime:

• ELEVENLABS_API_KEY (required)
  • Recommended: put it in ~/.openclaw/.env on the machine running the gateway.
• ELEVENLABS_VOICE_ID (optional but recommended)
  • If omitted, you must pass --voice-id when calling the script.

Example user prompts (to trigger this skill)

• "Reply with a voice note saying: …"
• "Generate audio (ElevenLabs) for: …"
• "Envíame una nota de voz que diga: …"

Workflow

1. Choose output path under the workspace, e.g. ./media/elevenlabs-tts/<timestamp>.mp3.
2. Run:

python3 {baseDir}/scripts/tts.py --text "..." --out "/abs/path/to/file.mp3" [--voice-id "..."]

3. Confirm with a short message including the output file path.

Defaults

• No voice is hardcoded. Set ELEVENLABS_VOICE_ID (recommended) or pass --voice-id.
• Model: eleven_multilingual_v2.