What problem does it solve? Adding a voiceover to a code-built video fails in predictable ways: batched text-to-speech silently drops lines, spoken audio runs about 50% longer than the written script, and mis-timed cuts produce audible clicks. This Skill encodes the measured workflow for writing narration for the ear, generating it within API quotas, and fitting the picture to the audio. ## Core Features & Use Cases - Quota-aware TTS generation: Batch two to four script sections per API call against a 10-requests-per-day free tier, with a JSON manifest driving one WAV per batch. - Transcription-based verification: Send finished audio back through a text model and compare it line by line against the script to catch missing lines, spoken stage directions, and silences that duration checks miss. - Render-time splitting and timing reconciliation: Play slices of batch files via Remotion trimBefore/trimAfter instead of cutting files, measure each section's real spoken length, and raise section durations and targetSeconds together so nothing clips. - Use Case: You have a 60-second explainer script and need a voice. This Skill batches the script into three API calls, verifies the transcripts, finds line boundaries with a silence scanner, reconciles the real ~95-second spoken length against the plan, and places each clip with 12-frame fade-ins and 15-frame fade-outs. ## Quick Start Ask the agent to add narration to your video using the adding-narration skill, starting from your approved script and confirming a TTS API key is available.
