What problem does it solve? Producing consistent, correctly timed voiceover for a product demo video is tedious: each storyboard section needs its own audio file, the narration length must match the planned picture duration, and pronunciation of product names often goes wrong. This Skill generates one measured audio file per storyboard section via the ElevenLabs MCP server so the editor can cut picture to voice. ## Core Features & Use Cases - Per-section narration generation: Creates one <section-id>.mp3 per storyboard section into demo/audio/, with voice, model, speed, stability, and similarity settings kept identical across sections. - Duration measurement and budget fitting: A probe hook measures each file against targetSeconds and guides rewriting the line, adjusting the budget, or tuning speed when a line runs more than 20% off. - Voice and model selection guidance: Reference docs cover audience-appropriate voice choice, ElevenLabs model trade-offs (multilingual_v2, v3, turbo, flash), and parameter ranges. - Pronunciation, captions, and music: Handles phonetic overrides for product names, optional word-level caption alignment via the forced-alignment REST endpoint, and optional ducked music beds. - Use Case: After a demo storyboard is approved, generate narration for all sections, listen to one section first to verify pronunciation of your product name, then regenerate any line whose measured length misses its target before reconciling the timeline. ## Quick Start Generate the voiceover audio for every section of my approved demo storyboard using the ElevenLabs voice we selected, then report any sections whose length misses the target duration.
