Tutorial voiceover generation with ElevenLabs

★ 6.5 · content

tutorial-voiceover-style governs the complete workflow in Claude Code for writing and generating voiceover strings in tutorial videos, covering work order, content rules, and ElevenLabs synthesis settings. The process enforces a strict sequence: lock all visual blocks (image/code/callout) in steps.json first, then write voiceover strings and run generate-tutorial-voiceover.ts — followed by two mandatory preview rounds before any render (first with audio only, then with SubtitleOverlay enabled). Content rules require that every key action, warning, and callout visible on screen gets mentioned in the narration, while padding phrases added purely to extend scene duration are explicitly forbidden. Speech synthesis uses the eleven_v3 model at 0.85 speed, configured via ELEVENLABS_SPEED in .env; individual clips can be regenerated without rebuilding the entire video by passing clip names to the script. The skill is essential for anyone editing intro/step/outro strings or running render:tutorial.