# Offset Frames identity revision

Final: [Does-AI-Help-You-Learn-Midcentury-Long.mp4](media/Does-AI-Help-You-Learn-Midcentury-Long.mp4), 625.5 seconds, 1920 × 1080, 24 fps, 15,012 frames.
SHA-256: `13441784b2a4d2848fc4ac00ccbcef259e8af3c2ec555960cf98dae6d87efae3`.

The studio now uses two offset film-frame outlines. The website masthead/footer, caption/tag ornaments, decorative scene marks and numeric plates were revised. The narration, story, timeline, captions, flags and research-related product marks retain their original contents.

- Full new encode decoded successfully; cadence, dimensions, duration and streams checked in [delivery record](review/delivery/delivery-check.json).
- Every old/new frame compared at matching index, at 640 × 360: [measurements](review/identity-revision/every-frame-before-after.csv). These numerical differences are descriptive, not an artistic similarity score.
- Every new frame was paired again with the original reference: [reference comparison](review/comparison/summary.json).
- All 65 shots manually checked in 17 before/after boards, plus selected native-resolution details. [Exact inspection coverage](review/visual-review.json). This is not manual inspection of every revised frame.
- Original audio preserved without re-encoding: all 29,322 AAC packets, timestamps and durations match. [Audio identity record](review/identity-revision/audio-identity.json). No new subjective listening is claimed.
- Ego Lite played the revised MP4, advanced through a middle segment and reached its ending without a media error. [Playback record](review/identity-revision/browser-playback.json) and [screen proof](review/identity-revision/browser-caption.png).
- Original timeline, narration text and both external subtitle files retain their previous hashes. The mux copies audio, embedded subtitles and chapter metadata from the retained original.

The previous release, QA and frame-review evidence remain in [before/](review/identity-revision/before/). Original full movies remain in `media/identity-revision-before/`. Their former visual review is not attributed to the new encode. The 30-second motion-review excerpt has also been regenerated. Chapter render caches and their original review records are historical; this revision renders the full film directly.

Run from the library root:

```sh
python3 scripts/audit_identity_revision.py --project productions/ai-learning-midcentury/editions/long
```

This revision audit verifies hash-bound retained evidence. It does not replace visual or audio review. The earlier `seal_review.py` and chapter-pixel audit apply to the archived original chapter assembly, not this directly rendered identity revision. No numerical claim of artistic equivalence is made.