# Static review of two selected gallery variants **Status: changes required.** Each selected rendering is moderately preferred to its earlier same-content comparator for the inspected incoming-image reveal. There is no overall winner between the two different content variants, no moving blind-test result, and no quality approval. ## Selected renderings and coverage | Content variant | Selected movie SHA-256 | Distinct inspected frame/content pairs | | --- | --- | --- | | Plant scene | `cb43238b58bf310234d51c8866e3540335f965e71c322a543fb8b0155b5e6946` | 59 | | Desk-holder scene | `e5cc0696c5aca1da029dcaa820c4d0bab0be2225886735febd9115a9c72c27ea` | 59 | Each movie contains 145 frames at 60/1 fps. **Selected-current coverage is 118 unique frame/content pairs, 59 per variant**, across the same 59 local time indices. Another **118 unique pairs belong to earlier comparison history**; they are not additional coverage of the selected renderings. Ten overviews and six individual native PNGs were viewed for the selected pair. The individual views repeat local indices 92, 125 and 143 already present in the overviews: 124 occurrences, but only 118 unique pairs. Comparison history has a separate 124 occurrences / 118 unique pairs. The original comparison's 236 unique pairs and 248 occurrences therefore reconcile exactly; neither total is a selected-current coverage claim. Each selected movie has 86 unseen target indices. All inspected output indices are target frames; output context count is zero. Source coverage is separately **160 distinct frames: 145 selected plus 15 context**, as detailed in `PUBLIC_SOURCE_OBSERVATIONS.md`. Reconciliation adds no image views, retiming, extraction, or coverage. ## Findings retained from the anonymous static comparison - At output local **92** / source retained **185**, both selected variants show a coherent green object in the incoming third card. The earlier same-content comparators showed a straight interior join between different object colors and contours. This supports the **moderate within-content preference**, not clearance of uninspected frames or a claim about code causes. - Local **125** / source **218** provides weaker supporting evidence: the selected fourth-card region already shows its blue setting rather than a residual warm portion of the earlier image state. - Local **143–144** / source **236–237** retains the gallery and terminal contraction. The original review found no decisive within-pair advantage at the resolved local-143 state. A small common terminal scale/framing difference was approximate, not an exact retiming diagnosis. Source cut **238** is outside the target and is not required inside either 145-frame output. ## Largest remaining gap and correction Both selected variants resolve into simplified illustrations with a repeated wall, geometric print and tabletop. The source resolves into distinct photographic scenes with material texture, lighting depth and different room details. This remains the **largest shared source-likeness gap**. Compare source 172–175 and 236–237 with inspected output local 79–80 and 143–144. For closer source likeness, use original, materially detailed images with distinct settings while retaining the white gallery, rounded tiles and staged reveals. Do not reuse the source lamp photographs. If illustration is intentional, record it as an accepted adaptation; it does not become a completed photographic match. The current evidence does not grant that acceptance. ## Subsequently disclosed implementation context After the source record and anonymous static findings were frozen, the project owner described a change to both base-card slots using their own SVGs and uniform-opacity detail instead of an advancing clip. This is attributed implementation context, **not independently inspected code and not the original visual proof**. It does not alter the recorded preference, explain the images with certainty, or increase coverage or approval. All approval, moving-blind, moving, audio, registration, quality, publication, Resolve and whole-film credits remain **0**. These documents have not been published. The frozen anonymous assessment remains unchanged.