ALLWAYS 056 / IN DIFFERENT HANDS / V02

35 seconds, 720x1248 portrait. Three pairs, six source clips: food preparation/deposition; mirror self-touch/barber service; covered fitting/exposed mechanism manually demonstrated. The final3.72 seconds of paired images are an editorial freeze, labelled in the film. No hidden mirroring, cropping, or speed changes.

Both independent critics passed visual pairing after the existing draft failed. One also compared the whole sequential control. Audio audition and native real-time screening remain unperformed. Original instrumental score has two separately saved panned stems. It is not recorded room sound. The film is an archive-led adaptation, not literal fulfillment of the chair/knot and50-second example. Full source contexts/rights are not independently established; access is not permission clearance.

The source/transform ledger, actual reports including failed baseline, and six exact source clips are bundled. The old failed cut is under baseline_failed/. The shared055/056 source kit is not modified by this dedicated kit.

REBUILD
Requires Python3, numpy, scipy, Pillow, ffmpeg and ffprobe. Unzip into empty directory, then run:
 python3 film56/make.py
 python3 film56/check.py
No source downloads are needed. The builder recreates paired film, panned sound stems, original mix and60-second silent sequential control. It writes a review-pending ledger; actual independent reports remain separate historical evidence.

A technical export check verified875 picture frames/35 seconds, both source identities per pair, full-frame aspect-preserving fits, no speed changes, the declared freeze durations, a60-second1500-frame sequential control, non-clipping original score, and exact equality of the first750 final picture frames to the film sent to critics. These checks do not establish artistic value.