Spatial-frequency hybrid illusion image prompt (two subjects in one)

2x2 grid, 16:9, AI inferred topics: Anchor: A single image that reads unmistakably as [HIGH-FREQUENCY SUBJECT (e.g., a wolf's face / a cathedral rose window / a coral reef)] when viewed large and close, and unmistakably as [LOW-FREQUENCY SUBJECT (e.g., a human skull / a clenched fist / a single closed eye)] when viewed as a thumbnail, squinted at, or seen from across a room ::4 Perceptual Mechanism: Spatial-frequency hybrid (Schyns–Oliva). The fine subject lives entirely in high-frequency detail - hair, edges, filigree, grain. The coarse subject lives entirely in low-frequency luminance blocking - broad light and dark masses. The two must share the same silhouette envelope so neither reading requires cropping ::4 Morphology: The coarse subject's dark masses are constructed FROM the fine subject's own features, never painted over them. No seams, no vignette, no compositing edge. Both readings occupy 100% of the frame ::3 Material Physics: [MEDIUM (e.g., charcoal on toned paper / silver gelatin photograph / dense engraving hatch)]. Monochrome or near-monochrome so luminance carries all the low-frequency information ::3 Illumination: Single hard source positioned so that cast shadow does double duty as low-frequency mass ::2 Viewing Protocol: Test at 1600px and at 120px. Both readings must survive. If either fails at its own scale, the image fails ::2 Render Stack: hybrid image, spatial frequency stimulus, [Chuck Close / Dalí / Salvador Dalí double image / Octavio Ocampo] ::1 Failure Condition: A cheap version shows both subjects at the same spatial frequency, so the eye sees a collage instead of a flip ::-1 Negative: [collage seams, transparency blending, ghosted overlay, second subject as outline, halo, vignette, colored fringing, either subject cropped] ::-1