Generate Voiceover

Turn a script into performance-ready speech, normally with ElevenLabs v3. This skill owns casting, direction, generation, review, and delivery; the ElevenLabs specialist contains deeper voice and emotion guidance.

Workflow

1. Confirm language, audience, speaker count, pronunciation, pace, mood, and target runtime.
2. Preserve the user's words. Ask before materially rewriting an approved script.
3. When the user asks for voice or audio examples, performance ideas, or a close starting point, use creativeclaw-find-examples with output_type: "audio"; load only the selected result. Do not search when the direction and script are already clear.
4. Call list_models({ modality: "speech" }), select speech/elevenlabs-v3 by default, then call get_model_params.
5. Choose a stock voice_id, or pass a saved character_id whose consented ElevenLabs clone should speak.
6. Add ElevenLabs v3 performance tags and punctuation deliberately. Keep agentic_prompting: false when exact script wording or control tags must survive unchanged.
7. Call generate_speech. Generate each speaker separately so casting, pacing, and revisions stay controllable.
8. Audition the result for pronunciation, emotional arc, pacing, clipping, and consistency. Regenerate only the weak section when practical.
9. Use merge_media when approved segments must become a single audio track or be muxed with video.

Conduct casting and review in the user's language. Preserve the supplied script and its writing system, and verify model or voice language support instead of translating unless the user asks.

Tool contract