docs: retime four films to 1.5x playback

[DivyamTalwar]

Summary

Replaces four of the thirty README films with 1.5x-speed versions. The originals ran 78–84 seconds each, which is a long hold for a reader skimming the gallery and longer than most social platforms show before a viewer drops.

Film	Before	After	Frames
01-first-run.mp4	83.63s	55.78s	2509
02-operators.mp4	77.97s	52.00s	2339
04-gates.mp4	78.67s	52.47s	2360
11-runs.mp4	83.63s	55.78s	2509

Why the frame count is unchanged

Every source is 30fps, and 30 × 1.5 = 45fps exactly. That means the retime is a pure timestamp remap — every frame carries through, with no drops and no duplicates, so there is no motion judder. A source at, say, 25fps would not divide cleanly and would have forced a compromise.

All four films are video-only, so no audio retime was required.

Encoding

CRF 18 from sources in the 230–400 kbps range, meaning the encoder over-allocates relative to the original and loses nothing visible. +faststart is set for web and social playback. Output is 1280×800 yuv420p, unchanged from source.

How to verify

ffprobe -v error -count_frames -select_streams v:0 \
  -show_entries stream=nb_read_frames,r_frame_rate,width,height \
  -show_entries format=duration docs/media/04-gates.mp4

Risk and rollback

Low. These are binary doc assets with no code path. Every README link resolves unchanged — the filenames are identical, only the content is faster. Poster frames and loop stills are unaffected, since retiming does not alter frame content. Rollback is git revert.

No README prose claims a duration for any of these four films. The only duration claims on the page are the 12.0s .png loops and film 27's on-screen exit 0 · 9.0s marker, neither of which is touched.

Scope

Four of thirty films. The remaining twenty-six still run at their original speed.

Verified

• ffprobe confirms 45fps, 1280×800, yuv420p and the expected frame count on all four
• check:qa-docs — pass
• check:brand — pass