ALLWAYS062 — IN USE — V02
60 seconds /960×540 /30fps. Chromatic archive study, plus picture-identical silent comparison.

EXPERIMENT
Yellow/gold links a pouring vessel and moving machinery. A yellow-and-blue sailing view makes the blue transition visible. Blue extends through a comparatively still prop arrangement, a child working at a table, and a canopy with red trim. Warm red then becomes visible through filling a glass and opening a labeled can. The closing view returns wider to the SAME pouring setup, where the yellow vessel, red contents and blue-cyan table covering share a practical event. It is a reprise, not a new pouring event or an unrevealed surprise after the red chapter has previewed it.

SOURCE / SCOPE
All used footage is from the original archive transfers of American Look (1958), PartsI–II, a promotional film. Settings differ within that source. This is an archival interpretation of prompt062, not the literal newly staged yellow-glove/repair/hidden-yellow sequence. Opening object is a yellow pouring vessel, not a glove. Machinery operates; it is not shown being repaired. The quiet blue prop view does not establish a hidden worker in the same shot. These differences must remain explicit rather than being treated as fulfilled literal proof conditions. The film does not assign universal emotional, cultural or danger meanings to hues. Its corporate source rhetoric is not erased merely by re-editing it.

COLOR
No hue replacement, grading, saturation boost, white balance, selective correction, recoloring or color-generated transitions. Original MP4s retained unchanged. The untagged SD transfers are decoded through one declared BT.601 limited-range RGB proxy; unknown original primaries are not reconstructed. Output uses an explicit BT.709 RGB→YUV matrix and709 tags. Colorimetric equivalence to original stock/emulsion or the user's physical display is not claimed. COLOR_PIPELINE.json preserves actual source stream metadata and assumptions. COLOR_REGIONS.json contains approximate manual rectangles, broad candidate hue masks, pixel area/location and limitations. Those are candidate filters, not object segmentation or an artistic pass score. Warm yellow/gold, blue/cyan and terracotta/burgundy red are families, not identical hue coordinates. The muted24–33second blue is a known weak point. Technical reference-frame checks compare prepared RGB to decoded export and report observed errors without declaring a universal perceptual tolerance.

TRANSFORMS
The main action body is uniformly fitted above the footer after a fixed content crop that removes most encoded black borders and slightly trims image edges. Opening and red-object crops, plus a gradual uniform-scale crop release in the final return, are retained frame by frame in COLOR_CONTEXT_TRANSFORMS.json. No warp/rotation/mirroring. Source playback durations retained while29.97fps transfers are normalized to30fps; no interpolated physical motion. Final3seconds of the shared setup hold its last recorded frame and use digital silence. This is an editorial still, not extended documentary observation or a synthetic physical event.

SOUND / COMPARISON
Main retains the supplied source audio on its own source intervals at unchanged gain; it may include production scoring/narration, not isolated synchronous contact recordings. It has not been listened to in this environment. No added score or new foley was created. No scrape/tap/contact or native-paced sound success is claimed. The PICTURE_ONLY comparison removes audio while copying the exact encoded video packets, so it tests the soundtrack's contribution without changing pictures. Its auditory comparison still needs a listener. Cards/terminal held-frame interval are digitally silent. SOUND_RECORD.json gives source ownership and exact intervals.

REPAIRS AND REVIEW
Two independent visual critics reviewed actual temporal sheets/native sampled frames, then actual source passages as needed. V01's red chapter weakened against blue tile and a distant golfer. V02 removes golf, begins red with a close view of the filling glass, and crops the opener around can/mechanism/hands with untouched colors. Actual review transcripts preserve initial failures and subsequent verdicts. Image-based criticism is not native real-time playback, auditory audition or target-display calibration. Scope remains qualified accordingly. No comparison-film superiority claim is made because the reference itself was not watched.

RECOVERY
Run python3 film62/build.py from the kit parent. Requires Python3, Pillow, numpy, ffmpeg/libx264, ffprobe, DejaVuSans. Source URLs/hashes and metadata in SOURCE_RECORD.json, exact timing in TIMELINE_V02.json, per-frame source/crop maps in COLOR_CONTEXT_TRANSFORMS.json. build_V01.py retains the first edit. verify.py validates export frames, decodes, video identity and reference pixels; color_records.py regenerates region diagnostics. An early incomplete MP4 attempt was discarded; explicit wait/probe gates now prevent accepting an incomplete export. The transient cause was not established. Unused32cached candidate clips and AmericanLook PartIII were inspected through saved contact sheets; their full unused media are omitted from the kit, with catalog URLs/metadata retained. Selected full source transfers are included.

RIGHTS / LINEAGE
Metadata rights labels and gaps are recorded, not independently audited. This is a private draft; no public release authorized or performed. Creator-upload text for Colors was inspected, not its complete film. New order and action choices are our method, not attributed to The Mercadantes. No Mozart reference soundtrack was reused.

FINAL REVIEW
Critics A and B both return a qualified visual PASS for V02 after the red repair. Aged/muted hue-family correspondence is accepted, not exact physical color equality. Muted blue24–33 is the weakest link. Final setting is a wider reprise, not a second pouring event or proof of social hospitality. Sound, native normal-speed reception and physical target-display calibration remain INCONCLUSIVE. The literal original glove/repair prototype was not produced.