voice_tts

What it does

• Converts text to speech via ElevenLabs.
• Saves the MP3 to scratch with a returned path.
• Use for podcast clips, voicemail drops, and demo voiceovers.

API spec

{
  "name": "voice_tts",
  "input_schema": {
    "properties": {
      "text": { "type": "string" },
      "voice_id": { "type": "string", "default": "default" },
      "model": { "type": "string", "default": "eleven_turbo_v2" }
    },
    "required": ["text"]
  }
}

Security

Goes through your ElevenLabs key (BYOK) or hosted account. Audit log captures text length and voice ID, not the text.

Used by

Gotchas

• Long inputs (>5000 chars) get chunked and concatenated.
• Voice cloning requires user consent verification.