Stylized isometric 3d diorama poster design

Please turn each photo I upload into an individual high-end design poster, no multi-image stitching, output each photo separately. Use a 3:4 vertical composition overall, with the top and bottom regions strictly 1:1 in height, each occupying 50% of the frame. 

The top half retains the original photo, maintaining the identity of the subject, structure, posture, realistic texture, natural lighting, and original color atmosphere, with only slight high-end color grading to give it the texture of an art magazine or exhibition image. To adapt to the frame, the environmental background can be naturally extended, but the subject must not be stretched, distorted, or changed.

The bottom half extracts only the most recognizable subject, outline, structure, posture, and narrative relationship of the photo; do not copy the photo completely, do not keep all objects, and do not redraw every object; actively delete most irrelevant backgrounds and details, keeping only the core information that best represents the original theme, and make a new artistic judgment. Even if the original composition is poor, the background is messy, or the subject is small, it must be re-directed into a valid work through deletion, reorganization, cropping, scale changes, and white space.

The aesthetic logic of the bottom half is reconstructed as Stylized Isometric 3D Architectural Diorama / Orthographic Axonometric 3D Render. Use an orthographic camera + three-quarter top-down axonometric perspective, approximately 30°–40° top-down and about 45° horizontal rotation, seeing the front, side, and top simultaneously, presenting a sense of space like a model, sand table, or miniature world. Do not make it a realistic photographic perspective, nor a pure map or elevation.

Reconstruct the subject into a stylized 3D miniature scene: with the subject as the core, retain only a few necessary environmental clues, such as terrain, roads, water, platforms, plants, buildings, or objects, to form a complete but highly summarized small world. The overall style is stylized 3D, not low-poly, and not realistic CG; emphasize regular geometry, moderate rounded corners, simplified materials, toy-like proportions, clean blocks, delicate layers, and soft lighting; it can have exquisite details, but cannot be complex or crowded.

The composition maintains a small-scale subject + massive amount of intentional white space. The miniature scene does not have to be centered and can be freely arranged according to the subject's direction, proportion, and visual center of gravity, and can be eccentric, edge-aligned, or partially cropped. White space is not an empty background but a part of the image, working with the subject to create a sense of breathing, space, and high-end order. Better to have less than too much.

The color scheme extracts 2–4 of the most vital and memorable colors from the top photo and reformulates them, moderately brightening, purifying, and removing gray to form a more bright, clean, and healing comprehensive color relationship. The overall tone remains fresh but not eye-piercing, warm but not cloying, and can naturally transform into refreshing color systems such as sky blue, lake cyan, vegetation green, cream yellow, apricot orange, coral pink, etc.; avoid gray-dirty, dark, old, fluorescent colors, and cheap plastic feel.

Text is used only as minimal editorial intervention, with no language restrictions. According to the subject, location, action, emotion, or metaphor, freely generate a small number of titles, short sentences, or notes, quietly landing on the white space area to form an exquisite graphic relationship with the image.

The overall presentation is a stylized axonometric 3D miniature scene, clear spatial relationships, exquisite model feel, vivid colors, massive amount of artistic white space, and editorial layout which together constitute a high-end visual effect. Avoid complete copying of the original image, complex stacking, realistic perspective, low-quality low-poly, cheap game feel, plastic feel, filling up the screen, and templated effects.