Invite d'image hybride à fréquence perceptuelle

4 images, 1:1, think hard, pick 4 stunning topics: <prompt> CORE_CONCEPT: draw A single-image spatial-frequency hybrid that reads as [HIGH_FREQUENCY_SUBJECT] up close and [LOW_FREQUENCY_SUBJECT] from afar. MEDIUM: [MEDIUM] IMAGE_LOGIC: Fine subject encoded only in high-frequency detail; coarse subject encoded only in low-frequency luminance masses. SURFACE: Monochrome or near-monochrome, tactile, richly textured, value-driven, no distracting color separation. ILLUMINATION: One hard directional source, placed so highlight/shadow architecture reinforces the distant-read image. COMPOSITION: Shared silhouette envelope, full-frame occupation, no cropping dependency, no dual-panel logic, no collage assembly. MOOD: Elegant, uncanny, intelligent, seamless, gallery-level, scientifically rigorous, and visually magical. </perceptual_frequency_hybrid_plate> <transformation> - Convert [HIGH_FREQUENCY_SUBJECT] into the close-range reading of the image. - Convert [LOW_FREQUENCY_SUBJECT] into the far-range reading of the image. - AI_INFER the strongest morphological correspondences between the two subjects so the distant image can be built from the close subject’s own forms. - Encode the close subject entirely through edge activity, texture density, filigree, hatch, grain, local contrast, and micro-structure. - Encode the distant subject entirely through broad value grouping, large light-dark shapes, cast shadow fields, and simplified mass logic. - Ensure both readings occupy the same silhouette envelope and use the entire frame. - Let the coarse subject’s cheekbones, eye sockets, jawline, fist mass, eyelid mass, skull dome, or other major structures emerge from the internal luminance choreography of the fine subject rather than from any overlay. - Build invisible transitions so the viewer cannot find the seam where one image becomes the other. </transformation> <semantic_inference> - AI_INFER which features of [HIGH_FREQUENCY_SUBJECT] can carry the most tonal weight for [LOW_FREQUENCY_SUBJECT]. - AI_INFER the optimal hard-light angle so shadow masses do the most perceptual work. - AI_INFER the correct balance of detail density vs tonal compression for [MEDIUM]. - AI_INFER how much contrast is needed so the thumbnail read survives without making the close read look artificially posterized. </semantic_inference> <detail_rules> - The close subject must be fully convincing on its own. - The distant subject must be unmistakable at a glance. - The two readings must not feel equally present at the same viewing scale. - The low-frequency image must not require blur effects, artificial glow, or faded transparency. - Every dark macro-mass must be earned by the detail structure of the fine subject. </detail_rules> <rendering_rules> - Favor value structure over color. - Keep tonal hierarchy clean and bold. - Use texture as information, not decoration. - Preserve exact edge discipline so close inspection remains rewarding. - Make the image suitable for gallery print, illusion study, conceptual editorial art, or museum poster reproduction. </rendering_rules> <finish> - True Schyns–Oliva-style hybrid logic. - Close-range intricacy, long-range inevitability. - Beautifully seamless dual readability. - Feels like an art-historical illusion piece and a perceptual science demonstration at once. </finish> <failure_conditions> - Fail if [LOW_FREQUENCY_SUBJECT] appears as an overlay. - Fail if [HIGH_FREQUENCY_SUBJECT] is sacrificed and stops reading clearly up close. - Fail if both subjects occupy the same spatial frequency band. - Fail if either reading depends on borders, vignette masking, or cropped framing. </failure_conditions> <negative> - collage seams - transparency blending - ghosted overlay - second subject as outline - halo reveal - vignette reveal - colored fringing - split-screen composition - either subject cropped </negative> }