---
name: sumi-elevenlabs
description: Help users securely connect ElevenLabs, choose narration voices/models, estimate character-based costs and save approved preferences for story videos; setup never authorizes paid takes.
---

# ElevenLabs narration setup

Read [the setup guide](references/setup.md). Reuse a connected voice connector or the bundled independent Python helper at `scripts/elevenlabs_voice.py`; it supports its documented platform credential routes. Start with saved `preferences`, then read-only connection/model/voice checks. Do not change a global voice for a project-only casting choice.

Explain character-based usage, real account/model/voice permissions and an estimate before any paid narration. Never collect API keys in chat, shell arguments, a project file or the ZIP. Guide restricted key creation and masked/OS-secure entry. Cloning needs consent and an eligible plan, not just a key. Offer a licensed library voice, personal recording or no narration if unavailable.

Save a reusable provider/voice/model only on the user's choice; retain receipts/take/alignment privately. Do not retry an uncertain paid generation: inspect its receipt/history. A voice can help identity; it cannot guarantee monetization. This skill does not sell or redistribute provider sound effects as a library. Sumi's bundled public sound bank is independent Python synthesis.