GPT Image 2.5

Use case: identity-preserve Asset type: character-consistency campaign scene Primary request: Generate a new photorealistic editorial scene of the exact blond male identity from Image 1 working late in a rainy architectural studio, leaning over a scale model and adjusting one small building piece with his right hand. Input images: Image 1 is the sole identity anchor. Preserve its face, hair, eyes, skin, stubble, body proportions, and age; do not copy the reference-sheet layout or black athletic outfit. Scene/backdrop: Contemporary architecture studio at night, tall rain-streaked windows, blurred blue city lights, warm task lamps, scale model and drafting materials with no readable writing. Subject and load-bearing identity proportions: One man only. Preserve the moderately long oval face, clearly longer than wide; broad cheekbones; firm jaw that tapers without becoming narrow; medium-square chin; straight balanced nose; brown eyes; light stubble. Preserve the high, wide mass of swept wavy dark-blond hair rather than flattening or shortening it. Preserve the athletic mesomorphic body: broad shoulders, developed but not bulky arms and chest, narrow waist, long balanced limbs. Age remains mid-to-late twenties. Wardrobe: Charcoal rolled-sleeve overshirt over a plain cream T-shirt; no branding. Composition/framing: Three-quarter waist-up editorial photograph, face unobscured, both forearms visible, right hand adjusting one model piece, left hand resting naturally on the table; entire hands remain in frame. Lighting/mood: Cool rainy-window ambience with warm amber task light; focused, quietly absorbed expression. Materials/textures: Real skin pores and stubble, individual hair strands, matte card model, worn wood table, rain on glass. Constraints: Identity fidelity and load-bearing proportions take priority over scene styling. Exactly one head, two eyes, two arms, two hands, five fingers per visible hand with natural joints; no fused or duplicated digits. No readable text, gibberish, logos, watermark, extra people, extra limbs, cropped hands, glamorized skin, or reference-sheet panels. Avoid: widened round face, narrow pinched jaw, flat or helmet-like hair, bodybuilder bulk, teenager appearance, beard, blue eyes, illustration, 3D render.

Use case: identity-preserve Asset type: character-consistency campaign scene Primary request: Generate a fresh candid photorealistic full-body street scene of the exact blond male identity from Image 1 walking through an early-morning flower market, carrying a plain canvas tote and smiling toward a flower stall. Preserve the same flower-market scene and identity intent while structurally removing visible footwear detail. Input images: Image 1 is the sole identity anchor. Preserve its face, hair, eyes, skin, stubble, body proportions, and age; do not copy the reference-sheet layout or black athletic outfit. Scene/backdrop: Open-air flower market with striped awnings, simple unlabeled buckets of flowers, pale wet stone street, softly blurred shoppers far behind, and subtle ground reflections. Subject and load-bearing identity proportions: One principal man. Preserve the moderately long oval face, clearly longer than wide; broad cheekbones; firm tapered jaw; medium-square chin; brown eyes; light stubble. Preserve the high, wide mass of swept wavy dark-blond hair. Preserve the athletic mesomorphic body: broad shoulders, developed but not bulky chest and arms, narrow waist, long balanced legs. Do not shorten the legs or enlarge the head. Wardrobe: Muted olive overshirt open over a plain white T-shirt and full-length dark tailored trousers. Both trouser legs use the same straight tailored cut and extend fully to the ground, with matching long cuffs that break cleanly and deeply over the feet. The cuffs fully cover all footwear on both feet. No shoe upper, collar, tongue, lacing, eyelets, ankle opening, or footwear construction is visible anywhere. At most, an identical simple flat dark toe/sole contact sliver may appear directly below each matching trouser cuff, without any upper detail. Unbranded cream canvas tote. Pose/activity: Natural measured walking step, relaxed smile, head turned slightly toward the flowers; right hand holds the tote handles, left arm swings naturally. Keep the step compact enough that both long trouser cuffs continue to cover the footwear completely. Composition/framing: Vertical full-body environmental portrait with both trouser hems, both ground-contact points, both arms, and both hands completely visible and contained. The full-length matching cuffs must obscure both shoes symmetrically. Wet-ground reflections may suggest two simple matching dark footwear shapes, but must not reveal or invent shoe details. Lighting/mood: Soft low morning sunlight, warm and spontaneous, realistic street color with restrained wet-stone reflections. Constraints: Identity fidelity and load-bearing proportions take priority. Exactly one principal head, two arms, two hands, two long trouser legs, two fully covered feet, and two matching trouser cuffs; five fingers per visible hand with natural joints; normal gait and weight distribution. Footwear must be fully concealed behind the cuffs on both feet. No visible shoe upper, collar, shaft, tongue, lace, eyelet, panel seam, logo, or mismatched footwear detail. Background shoppers remain indistinct and do not resemble the subject. No readable text, gibberish, logos, watermark, duplicated subject, fused limbs, extra fingers, cropped hems, cropped ground-contact points, or reference-sheet panels. Avoid: visible sneakers, visible shoe construction, exposed ankles, raised trouser cuffs, different hem lengths, asymmetric cuff breaks, round widened face, narrow pointed chin, flat hair, bodybuilder bulk, short stocky legs, beard, blue eyes, fashion-editorial airbrushing, illustration, 3D render.

Use case: identity-preserve Asset type: character-consistency campaign scene Primary request: Generate a new photorealistic lifestyle scene of the exact blond female identity from Image 1 browsing a quiet vintage record store and lifting one unbranded record sleeve from a wooden bin. Input images: Image 1 is the sole identity anchor. Preserve its face, chin-length bob, eyes, skin, body proportions, and age; do not copy the reference-sheet layout or black athletic outfit. Scene/backdrop: Small vintage record shop with wooden bins, softly blurred shelves, warm front-window light, and abstract unlettered sleeve art. Subject and load-bearing identity proportions: One woman only. Preserve the softly heart-oval face, only modestly longer than wide; broad cheekbones; compact lower face; small tapered chin; brown almond-shaped eyes; straight small nose; full but natural lips. Preserve the rounded chin-length blonde bob: side part, smooth swept fringe, outer hair mass wider than the jaw, ends curving inward at jaw level. Preserve the compact athletic hourglass build with moderate shoulders, defined waist, balanced hips, and neither elongated nor extremely thin limbs. Wardrobe: Simple rust-colored knit top with elbow-length sleeves and high-waisted dark denim; no branding or jewelry that obscures identity. Pose/activity: Three-quarter mid-length view, both hands naturally holding the opposite edges of one record sleeve, small pleased smile, eyes glancing at the sleeve. Lighting/mood: Warm diffused window light, intimate and quietly delighted. Constraints: Identity fidelity and load-bearing proportions take priority. Exactly one head, two arms, two hands, five fingers per visible hand; clean sleeve geometry and natural grip. No readable text, gibberish, logos, watermark, extra people in focus, extra limbs, fused fingers, cropped hands, photorealistic identity drift, or reference-sheet panels. Avoid: long narrow face, enlarged chin, long hair, pixie cut, flattened bob, oversized hair helmet, tall runway-thin body, exaggerated curves, blue eyes, heavy makeup, illustration, 3D render.

Use case: identity-preserve Asset type: character-consistency campaign scene Primary request: Generate a new natural photorealistic portrait of the exact red-haired female identity from Image 1 tending seedlings in a misty glasshouse, steadying one terracotta pot and watering it with a small metal can. Input images: Image 1 is the sole identity anchor. Preserve its face, freckles, eyes, hair, body proportions, and age; do not copy the reference-sheet layout or black athletic outfit. Scene/backdrop: Humid glasshouse with fogged panes, wooden potting bench, restrained green seedlings, terracotta pots, and no signage. Subject and load-bearing identity proportions: One woman only. Preserve the narrow long heart-oval face, clearly longer than wide; high cheekbones; slim jaw; small tapered chin; green-hazel eyes; dense natural freckles across nose, cheeks, forehead, arms, and hands. Preserve the tall loose copper-red bun that adds substantial vertical hair mass above the skull, with fine wispy tendrils at both temples and nape. Preserve the slim long-limbed athletic hourglass build with narrow shoulders, defined waist, balanced hips, and long forearms and legs without emaciation. Wardrobe: Forest-green fitted long-sleeve henley and practical tan trousers; no branding. Pose/activity: Three-quarter knee-up view, left hand steadying one small pot, right hand holding a small watering can, natural wrists and finger placement, gentle closed-mouth smile. Lighting/mood: Soft overcast daylight diffused through wet glass; quiet, hopeful, tactile. Constraints: Identity fidelity and load-bearing proportions take priority. Exactly one head, two arms, two hands; five fingers per visible hand with natural joints; watering-can handle and spout must be coherent. No readable text, gibberish, logos, watermark, extra people, extra limbs, fused fingers, cropped hands, missing freckles, or reference-sheet panels. Avoid: round short face, wide heavy jaw, low compact bun, straight loose hair, pale unfreckled skin, stocky build, exaggerated thinness, brown eyes, glam makeup, illustration, 3D render.

Use case: identity-preserve Asset type: character-consistency campaign scene Primary request: Generate a new natural photorealistic scene of the exact merle French bulldog identity from Image 1 standing alert on wet coastal rocks and investigating a small tidepool in soft morning mist. Input images: Image 1 is the sole identity anchor. Preserve its head, ears, muzzle, eyes, merle pattern language, body proportions, paws, and age; do not copy the reference-sheet layout or gray studio background. Scene/backdrop: Quiet rocky coast at low tide, one small reflective tidepool, wet charcoal stones, faint sea foam and misty horizon; no people or other animals. Subject and load-bearing identity proportions: One dog only. Preserve the broad nearly square head; tall upright bat ears set wide apart with rounded tips; extremely short broad muzzle approximately one quarter of head length; broad dark nose; pale blue-gray eyes set wide; cream blaze on lower face and chest. Preserve the compact cobby muscular torso, wide chest, short sturdy legs, slightly longer body than tall, broad paws, and very short tail. Preserve blue-gray and warm-cream merle patches as irregular mottling rather than stripes or spots. Pose/activity: Natural four-square standing pose with head lowered slightly toward the pool, ears upright, curious calm expression, all four legs and paws visible and bearing plausible weight. Composition/framing: Low eye-level landscape photograph, full dog contained with clear separation between all four legs; reflection may appear only inside the tidepool and must not resemble a second dog. Lighting/mood: Diffused cool morning light, soft mist, quiet curiosity. Materials/textures: Short realistic fur, wet nose, detailed paw pads/claws where visible, water ripples, slick stone. Constraints: Identity fidelity and load-bearing proportions take priority. Exactly one dog, one head, two ears, four legs, four paws; anatomically coherent joints and toes. No collar, clothing, readable text, gibberish, logos, watermark, extra animal, duplicated face, extra leg, fused paw, long muzzle, floppy ears, long tail, cropped paws, or reference-sheet panels. Avoid: tall lean terrier proportions, narrow head, long snout, low ear set, blue coat without cream, dalmatian spots, cartoon style, 3D render.

Use case: style-transfer Asset type: identity-across-media campaign hero Primary request: Re-render the exact flat-vector-style Sema head from Images 1 and 2 as a handcrafted glazed ceramic wall portrait photographed as a physical object. Change the medium only; preserve the mascot identity and proportions. Input images: Image 1 is the primary silhouette, viewing-angle, hair-sweep, face-inset, line-placement, and palette anchor. Image 2 is the primary eye, lash, mouth, and expression-language anchor. Scene/backdrop: Warm charcoal mineral-plaster wall, subtle museum plinth edge, no other objects. Subject and load-bearing identity proportions: Exactly one Sema head, no body. Preserve the near-circular horizontal oval silhouette, slightly wider than tall; compact bob ending in two rounded lower side lobes; heavy golden fringe beginning near top-center and sweeping diagonally across the forehead to the lower-right cheek; shorter pointed lock on the left. Preserve the warm-ivory face inset centered low inside the hair mass at roughly two-thirds of total head width, with wide soft cheeks and a small rounded chin. Preserve evenly spaced black oval eyes at mid-face, outer winged lashes, no nose, and a tiny curved approving mouth well below the eyes. Medium/material: Real handmade glazed ceramic in shallow relief, cream face glaze, honey-gold and ochre hair glazes, glossy near-black outlines/features. Slight glaze pooling and kiln variation, but contours remain crisp and readable. Composition/framing: Square studio hero, object fully contained with generous wall margin, slight three-quarter physical depth, no crop. Lighting/mood: Soft directional museum light, gentle physical shadow, refined and tactile. Constraints: Medium translation must not alter identity geometry. No body, neck, ears, nose, extra face, duplicated object, flattened printed illustration, text, labels, logos, watermark, chipped missing features, or cropped silhouette. Avoid: taller egg-shaped head, narrow face inset, centered curtain bangs, reversed hair sweep, oversized eyes, realistic human face, emoji redesign, plastic toy, vector-flat rendering.

Use case: style-transfer Asset type: identity-across-media material study Primary request: Create one polished 2-by-2 study sheet showing the exact same Sema head from Images 1 and 2 physically re-rendered in four unmistakably different media: needle-felted wool, glazed ceramic relief, carved and painted wood, and translucent cast resin. Preserve identity geometry in every cell; change material only. Input images: Image 1 is the primary silhouette, viewing-angle, hair-sweep, face-inset, line-placement, and palette anchor. Image 2 is the primary eye, lash, mouth, and expression-language anchor. Scene/backdrop: Four equal charcoal specimen cells separated by thin warm-gray rules; no labels or captions. Subject and load-bearing identity proportions: The same one Sema head in every cell, no body. Preserve the near-circular horizontal oval silhouette, slightly wider than tall; compact bob ending in two rounded lower side lobes; heavy golden fringe beginning near top-center and sweeping diagonally across the forehead to the lower-right cheek; shorter pointed lock on the left. Preserve the warm-ivory face inset centered low inside the hair mass at roughly two-thirds of total head width, wide soft cheeks, small rounded chin, evenly spaced black oval eyes at mid-face, outer winged lashes, no nose, and tiny curved mouth. Cell media: top-left needle-felted wool with visible fibers and stitched black features; top-right glossy glazed ceramic relief with glaze pooling; bottom-left carved and painted wood with shallow tool marks and visible grain; bottom-right translucent cast resin with suspended golden pigment and embedded opaque face/features. All retain the same cream, gold, ochre, and black palette. Composition/framing: Exact 2-by-2 grid, one fully contained head per cell, matching scale, viewing angle, face placement, and expression, generous padding, no overlap. Lighting/mood: Even museum specimen lighting tailored subtly to reveal each material. Constraints: Identity proportions must match across all four cells and the references. Exactly four heads total. No body, neck, ears, nose, extra face, duplicated cell, text, labels, logos, watermark, cropped silhouette, or medium ambiguity. Avoid: taller oval heads, narrow face insets, reversed or centered fringe, realistic human faces, four color variants of the same material, flat printed vector copies, inconsistent eye spacing or mouth placement.

A bright beverage advertisement photograph of a chilled can of orange sparkling soda, standing at a slight tilt on a smooth pale surface, shot on a 3:4 vertical canvas at close product distance. COMPOSITION The can is the largest and most central element, occupying about 55% of the frame height, its base resting just below the lower third. A single thin orange slice curls behind the can's upper third, its circle echoing the can's rim. One narrow arc of liquid rises from the surface behind the can and sweeps around its right side at the can's mid-height, then falls out of frame. The arc passes BEHIND the can and never crosses the label. The composition is built on three circles — the slice, the rim, the liquid arc — held on one vertical axis. Empty space falls to the lower left. Nothing touches the frame edges. SETTING AND LIGHT Background: a seamless warm cream backdrop with a soft apricot falloff toward the lower right, no props, no horizon. Lighting: one large soft source behind the can at the upper left, acting as a rim light that passes through the orange slice and glows through the liquid arc; one broad white bounce card in front at low intensity to keep the label legible. Contact shadow: one tight soft shadow directly beneath the can base. No coloured gels, no second light, no reflections of studio equipment. MATERIALS Cold aluminium can with a satin brushed finish and a fine bead-blasted sheen along the rim; heavy condensation across the cold metal, with droplets varying from fine mist to a few larger runs, all catching the back light. Orange slice: translucent flesh with visible radial segments and a pith ring, the back light glowing through it so the pulp reads as wet rather than as coloured glass. Liquid: thin, glossy, opaque orange with a defined meniscus, heavy enough to hold a smooth curve, catching a bright highlight line along its top edge. LABEL AND TEXT The label is white with a matte paper texture, wrapped cleanly and registered straight to the can's vertical axis. It carries exactly this text, once each: the brand word "ZEST" in a bold condensed sans on the upper third, "ORANGE" beneath it in a smaller light sans, and "330 ML" at the bottom of the label. No other words, numbers, marks or symbols appear anywhere in the frame. Brand text must be crisp, correctly kerned, unobstructed and readable at thumbnail size. CONSTRAINTS One liquid arc only. No scattered fruit pieces, no splashes, no droplets scattered through the air. The can's silhouette, the label's proportions and the label's registration must stay unchanged. Do not add a second can, a glass, ice cubes, leaves or a straw. No invented logos, no badges, no awards, no "new" or "best" flashes, no watermark. Avoid plastic-looking CGI, avoid flat airless lighting, avoid over-sharpened edges.