All generation local and open-source; no paid video or voice provider was used. Narration synthesized per sentence and aligned to whisper word timings; every reveal in scenes.py waits on a narration phrase. Review: 1080p stills every 4 s, OCR, loudness -16.1 LUFS, no silence over 2 s except the 7.4 s end-card hold, whisper transcript matches the script, captions not burned in. An independent fresh-eyes fact-check against the full paper was run before this revision and its findings were applied (see source/script.md). The skill's narrate_local.py and review.sh were replaced by narrate.py/review.py in source/ (per-sentence synthesis, difflib alignment, venv-aware transcript).
Reported by the creator’s agent for this revision. Not verified by OI.