| name | elevenlabs-voice-clone |
| description | Voice cloning and conversational voice configuration using ElevenLabs for reusable, branded AI phone voices. |
| category | Communication |
| roles | ["sparky-chief-product-quality-officer","product-manager","backend-architect","mcp-integration-engineer"] |
ElevenLabs Voice Clone
Status: READY FOR INTEGRATION
Purpose
Create and manage a branded cloned voice for AI phone interactions without hardcoding voice assets or mixing voice data into the main delivery runtime.
Requirements
- Use ElevenLabs API credentials from environment variables only.
- Keep voice-clone source recordings out of git.
- Store only the resulting
voice_id in runtime configuration.
- Separate voice creation from call-serving runtime so production calls can keep working while the voice evolves.
Configuration Template
ELEVENLABS_API_KEY=your_api_key
ELEVENLABS_VOICE_ID=voice_abc123
ELEVENLABS_AGENT_ID=agent_abc123
ELEVENLABS_MODEL_ID=eleven_v3_conversational
ELEVENLABS_OUTPUT_FORMAT=ulaw_8000
Voice Clone Workflow
- Capture clean sample audio outside the repo.
- Upload samples to create a cloned voice.
- Record the resulting
voice_id in a secret or environment variable.
- Bind the voice to the phone-answering runtime.
- Run a smoke call and judge intelligibility, latency, and brand fit.
Safety Rules
- Do not impersonate real people without explicit approval.
- Keep a neutral fallback voice available.
- Treat source recordings as sensitive assets.
- Prefer low-latency telephony-compatible output formats for live calls.
Evidence
- Voice clone created successfully.
voice_id is configured in the runtime.- Test call audio is intelligible and stable.
- Fallback voice path is documented.
