Back to gallery

Real or Not

GPT IMAGE 2.5 → GPT IMAGE 2.5 SUNBURST → SEEDANCE 2.5FORMAT 4:3 · 30 S · 1080P · PLATFORM ARCADSPhoto

Find an accomplice

SEEDANCE 2.5 · VIDEO PROMPT · 30 S · 4:3 · 1080P

FORMAT: 30-second cinematic multi-shot sequence, 4:3, 1080p, native dialogue and sound. Hard cuts between shots, real-time action, no slow motion. STYLE: A real police-station scene filmed on an iPhone, with the texture of an ARRI Alexa and ARRI Signature Prime lenses: low contrast, lifted blacks, soft highlight roll-off, creamy falloff, muted desaturated grade, fine grain. Handheld micro-shake, no smooth gimbal, except the lineup tracking shot. Lineup room: cold flat fluorescent light. Observation room: dark, lit only through the glass. REFERENCES: @Image 1: LOCATION, BLOCKING AND GRADE. The observation room, the one-way glass, the lineup wall, the four suspects holding cards 1 to 4, the old woman and the detective in the foreground. Not a first frame: start with gentle movement. @Image 2: SUSPECT 1, the platinum-ponytail woman. Character sheet: face, hair, outfit, silver hoops, nose stud. @Image 3: SUSPECT 2, the young man with black glasses. Character sheet. @Image 4: SUSPECT 3, the moustached man with cap and headphones. Character sheet. @Image 5: SUSPECT 4, the curly-haired man in the burgundy leather jacket. Character sheet: outfit and identity. @Image 6: THE OLD WOMAN. Character sheet. @Image 7: THE DETECTIVE. Character sheet. @Image 8: THE BOOK, with its deep red paper belly band. Use the front cover exactly as shown. @Image 9: SUSPECT 4 FACE SUPPORT. A real photo of the same man: use only his face to sharpen the likeness, bone structure, eyes, nose, moustache and goatee. Ignore clothing, room and selfie light. Character sheets: restore each face precisely in every shot; ignore grids, labels and grey backgrounds. VIDEO PROMPT: 0.0 to 2.0s. Exactly the framing of @Image 1: behind the old woman and the detective, looking through the glass at the four suspects. The camera eases in slowly. 2.0 to 4.0s. Close-up of the detective from the glass side, the lineup reflected in his lenses. He points his notebook at the glass, dry and tired. {So, ma'am. Which ones are AI?} 4.0 to 5.5s. Close-up of the old woman leaning toward the glass, eyes narrowing, studying each face. 5.5 to 16.5s. Inside the lineup room, one continuous smooth lateral tracking shot on a dolly at chest height, gliding left to right without ever stopping, pushing in and drifting back out as it passes each suspect. 5.5 to 8.5s. Suspect 1 in close-up, chewing gum slowly, bored, eyes half-lidded, unbothered. {She will never figure it out.} 8.5 to 11.0s. The camera drifts in to an extreme close-up of Suspect 2's forehead: a single drop of sweat runs from his hairline down past his glasses. It eases back to the wide sliding angle; he lifts his chin, trying to stay cool, eyes flicking nervously. {It is pretty obvious I am real.} 11.0 to 14.0s. Suspect 3, jittery: sniffing, rubbing his nose, bouncing on his heels, scratching his neck, eyes darting, card trembling. He repeats fast under his breath. {Not me. Not me. Not me. Not me.} 14.0 to 16.5s. The camera reaches Suspect 4, relaxed, leaning slightly back. It pushes in very close to his mouth as one corner slowly pulls into a cool half smile. The track ends here. 16.5 to 18.0s. Hard cut: close-up of the old woman, quiet and certain. {One, two and three.} 18.0 to 20.5s. Lineup room, wide. A steel side door opens; the detective and two uniformed officers walk in. The detective stops at Suspect 4, tilting his head to the door. {Number four. You are free to go.} 20.5 to 23.0s. Officers cuff Suspects 1, 2 and 3: the woman rolls her eyes chewing, the young man stares in shock, the moustached man mouths silently. Suspect 4 drops his card and walks out. 23.0 to 26.0s. A police-station corridor at a property counter. A bored officer slides him a canvas tote bag. {Your belongings.} He takes it with a small nod. 26.0 to 30.0s. Over-the-shoulder from behind his right shoulder at head height as he walks down the corridor: he pulls the black hardcover book from @Image 8 out of the tote and holds it in front of his chest with both hands, cover up, looking down at it. The camera pushes in over his shoulder; he slows and stops. The cover reads exactly: ULTIMATE CINEMATIC REALISM BLUEPRINT, the arcads logo, six portraits numbered 01 to 06, and the red band: FREE TRIAL · 1,000 FREE CREDITS, FOR ALL NEW ARCADS CUSTOMERS, as on @Image 8. Final 1.5 seconds hold on the cover, sharp. AUDIO: Diegetic sound effects only, no music. Fluorescent hum. Gum chewing on Suspect 1. Sniffing and shuffling feet on Suspect 3. <steel door buzzer and clunk> at 18.0s. <handcuffs clicking shut> at 21.0s, 21.6s and 22.2s. Footsteps on linoleum. <canvas tote sliding on the counter> at 23.5s. Soft rustle of the book from 26.5s. VOICES: DETECTIVE: male, mid-fifties, low dry American English voice, slightly gravelly, weary and unhurried. OLD WOMAN: female, seventies, soft American English voice, shaky but firm. SUSPECT 1: female, early twenties, flat bored American English voice with vocal fry, gum in her mouth, smug. SUSPECT 2: male, mid-twenties, American English, trying to sound calm, his voice cracking slightly. SUSPECT 3: male, thirties, American English, fast, breathy, high-strung, muttering. OFFICER: male, forties, flat bored American English voice. SUSPECT 4 never speaks. Each line is spoken once, by its own character, in the shot where it appears. CONSISTENCY: Every face matches its character sheet in every shot; Suspect 4 also matches @Image 9. Card numbers never change. Suspect 1 chews gum throughout. Suspect 2 keeps his glasses on. Suspect 3 keeps cap, headphones and glasses on. Suspect 4 never wears sunglasses. The tracking shot moves left to right only and never stops. The book cover is identical to @Image 8. NEGATIVE PROMPT: music, extra spoken words, narration of the prompt, subtitles, captions, changed faces, swapped card numbers, sunglasses on Suspect 4, garbled book title or band text, stuttering camera, character sheet grids or labels, high contrast, crushed blacks, slow motion, watermark

Other prompts in this tutorial

Prompt 1

GPT IMAGE 2.5 · CHARACTER CHART · SUSPECT 1 { "task": "Photorealistic fashion character reference sheet for AI film production, assembled from real photographs taken in one studio session.", "references": { "image_1": "IDENTITY ONLY. This exact young woman: her face, bone structure, pale blue-grey eyes, thick dark brushed-up brows, straight nose, full soft pink lips, fair skin with natural freckles across the nose, cheeks and shoulders, and her platinum ash-blonde hair pulled back tight into a high sleek ponytail with a few loose flyaway strands, the small dark fine-line tattoo behind her ear on the side of the neck. NEVER copy the bare shoulder or the shoulder tattoo exposure, the dark background or the harsh hard sunlight.", "image_2": "WARDROBE ONLY. Copy the outfit exactly: oversized chunky-knit burgundy crewneck sweater with ribbed cuffs and hem, a long cream chunky-knit scarf draped around the neck with both ends hanging down the front, a light-wash blue denim mini skirt, sheer black tights, white ribbed knee socks folded over the top of black leather knee-high biker boots with buckle straps, and a black slouchy leather shoulder bag with silver hardware. NEVER copy her face, her long dark curly hair, the phone, the mirror, the bedroom or the selfie pose.", "image_3": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "jewellery": "New additions not in either reference: small polished silver hoop earrings in both earlobes plus a tiny silver stud in the upper helix of her left ear, and a delicate, elegant tiny silver stud piercing on the left side of her nose, minimal and chic. Visible in every panel where her face or ears show.", "layout": { "top_row": [ "Full-body front, relaxed stance, one hand at her side, bag on her left shoulder, cool neutral expression, eyes to lens", "Full-body left profile, chin slightly raised, ponytail clearly visible, scarf ends hanging forward, boots and folded socks in clean silhouette", "Full-body back view, sleek platinum ponytail, back of the burgundy sweater, scarf around the neck, bag strap over the shoulder", "Large head-and-shoulders close-up, front, looking straight into the lens, freckles, hoop earrings and nose stud sharp, scarf around the neck", "Three-quarter close-up turned to her right, looking back over her shoulder at the camera like Image 1, ear with hoop and helix stud and the small neck tattoo visible, nose stud catching a soft highlight" ], "bottom_row_labels_exact": ["PLATINUM PONYTAIL", "SILVER JEWELLERY", "CHUNKY KNIT", "CREAM SCARF", "DENIM SKIRT", "BIKER BOOTS", "COLOR PALETTE"], "palette_swatches": ["burgundy", "cream", "light denim blue", "black", "platinum blonde", "silver"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the boots, light grey seamless paper with a faint floor curve.", "realism": "Unretouched skin with visible pores, real freckles and fine peach fuzz, natural lip texture, real flyaway hairs at the hairline. Knit texture with visible stitches and a little pilling, denim weave and fading, matte sheerness of the tights showing skin tone, creased leather on the boots and bag, small reflections on the silver jewellery. Use case: professional casting and wardrobe board.", "constraints": "Same woman in every panel, her face comes only from Image 1. Platinum ponytail in every panel, never long dark curls. Earrings and nose stud consistent in every panel. No phone, no mirror, no bedroom. No beauty retouching, no plastic skin, no extra people, no watermark, no text other than the seven labels." }

Prompt 2

GPT IMAGE 2.5 · CHARACTER CHART · SUSPECT 2 { "task": "Photorealistic fashion character reference sheet for AI film production, assembled from real photographs taken in one studio session.", "references": { "image_1": "IDENTITY ONLY. This exact young man: his face, bone structure, sharp jawline, dark brown almond eyes, straight dark brows, straight nose, full lips, light stubble on the upper lip and chin, warm light skin with visible pores, his messy black middle-parted hair falling in loose strands over the forehead and around the eyes, and the small round pearl stud earrings in both ears. NEVER copy the bare chest, the white backdrop or the wet skin sheen.", "image_2": "WARDROBE ONLY. Copy the outfit exactly: an off-white cream waffle-knit long-sleeve henley with a three-button placket, worn untucked and loose, sleeves pushed up slightly on the forearms; navy pinstripe wide-leg pleated trousers with a high rise and a long full break; cream and grey suede low-top sneakers with gum-coloured soles; black rectangular acetate glasses; white wired earphones hanging from his ears; a thin silver chain necklace; a silver metal-strap wristwatch on his right wrist; a couple of thin silver rings; a black nylon crossbody shoulder bag on his left shoulder. NEVER copy his face, his wavy dark hair, the parking lot or the street.", "image_3": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "layout": { "top_row": [ "Full-body front, relaxed stance, left hand in trouser pocket, bag strap over the shoulder, calm neutral expression, glasses on, eyes to lens", "Full-body left profile, mid-stride walking pose, the wide trouser legs swinging, earphone wire visible", "Full-body back view, back of the henley, bag hanging on his left side, trouser hems pooling over the sneakers", "Large head-and-shoulders close-up, front, glasses on, loose black strands over the forehead, pearl stud and earphone visible, henley placket and silver chain sharp", "Three-quarter close-up turned slightly to his left, glasses on, looking just past the camera, profile of the jaw, pearl earring and earphone wire catching the light" ], "bottom_row_labels_exact": ["MIDDLE-PART HAIR", "WAFFLE HENLEY", "PINSTRIPE TROUSERS", "SUEDE SNEAKERS", "GLASSES & PEARLS", "SKIN TEXTURE", "COLOR PALETTE"], "palette_swatches": ["cream", "navy pinstripe", "black", "silver", "gum brown", "warm skin tone"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the sneakers, light grey seamless paper with a faint floor curve.", "realism": "Unretouched skin with visible pores, faint stubble texture, natural lip texture, real stray hairs. Waffle knit with visible grid texture and slight stretch at the elbows, fine pinstripes that stay straight across pleats and folds, suede nap on the sneakers, small reflections on the lenses, pearls and silver. Use case: professional casting and wardrobe board.", "constraints": "Same man in every panel, his face and hair come only from Image 1. Glasses on in every panel. No phone, no street, no extra people, no beauty retouching, no plastic skin, no logos, no watermark, no text other than the seven labels." }

Prompt 3

GPT IMAGE 2.5 · CHARACTER CHART · SUSPECT 3 { "task": "Photorealistic fashion character reference sheet for AI film production, assembled from real photographs taken in one studio session.", "references": { "image_1": "IDENTITY ONLY. This exact man: heavyset build, broad round face with full cheeks and a soft double chin, hazel-brown eyes under thick dark brows, a rounded nose, a thick dark brown waxed handlebar moustache with curled tips, light stubble on the chin and jaw, fair skin with a ruddy flush on the cheeks and nose and a few freckles, and his shoulder-length messy wavy dark brown hair. NEVER copy the charcoal T-shirt or the white wall.", "image_2": "WARDROBE ONLY. Copy the outfit exactly: a teal-green baseball cap with a small red embroidered patch on the front, worn forward; black over-ear wireless headphones worn over the cap; thin round black wire glasses; an oversized boxy white cotton T-shirt with dropped shoulders and a colourful psychedelic illustrated graphic on the chest (cartoon flowers, small creatures and a pink figure, in bright pinks, blues, greens and yellows); dark olive textured ripstop shorts above the knee with a small yellow woven label tab on the left pocket; a black smartwatch; white ribbed crew socks pulled up; chunky beige-cream caged hiking-style sneakers with thick lugged soles. Keep the T-shirt graphic as an original illustration with no readable text. NEVER copy his face, the phone, the mirror, the hallway or the watermark.", "image_3": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "hair_with_cap": "His wavy dark brown hair spills out from under the cap and the headphones at the sides and back, falling to the shoulders.", "layout": { "top_row": [ "Full-body front, feet apart, one hand in his shorts pocket, cap, headphones and glasses on, deadpan expression, eyes to lens", "Full-body left profile, belly in silhouette, hair falling behind the headphone band, chunky sneakers in clean side view", "Full-body back view, back of the cap and headphone band, wavy hair over the collar, plain back of the white tee, olive shorts", "Large head-and-shoulders close-up, front, cap, headphones and round glasses on, handlebar moustache sharp, the top of the T-shirt graphic visible", "Three-quarter close-up turned to his right, adjusting the headphone ear cup with one hand, moustache curl and cap patch catching the light" ], "bottom_row_labels_exact": ["CAP & HEADPHONES", "HANDLEBAR MOUSTACHE", "GRAPHIC TEE", "OLIVE SHORTS", "CHUNKY SNEAKERS", "SKIN TEXTURE", "COLOR PALETTE"], "palette_swatches": ["teal green", "white", "olive drab", "cream beige", "black", "hot pink"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the sneakers, light grey seamless paper with a faint floor curve.", "realism": "Unretouched skin with visible pores, ruddy cheeks, slight oily shine on the nose and forehead, individual moustache hairs and wax sheen on the curled tips, frizzy flyaway strands in the wavy hair. Real body weight: the tee draping over the belly, natural creases at the sides. Cotton tee with slightly faded print texture, ripstop grid on the shorts, mesh and rubber on the sneakers, soft leatherette on the headphone cushions, small reflections on the glasses. Use case: professional casting and wardrobe board.", "constraints": "Same man in every panel, his face, hair and moustache come only from Image 1. Cap, headphones and glasses on in every panel. No phone, no mirror, no extra people, no beauty retouching, no plastic skin, no readable logos, no watermark, no text other than the seven labels." }

Prompt 4

GPT IMAGE 2.5 · CHARACTER CHART · THE DETECTIVE { "task": "Photorealistic character reference sheet for AI film production, assembled from real photographs taken in one studio session. The character is a seasoned city detective.", "references": { "image_1": "IDENTITY ONLY, THE MAN IN THE FOREGROUND. Take only the man in front with his arms crossed: his lean angular face, high cheekbones, deep-set light eyes, straight nose, thin lips, short grey-flecked dark brown hair, short grey and brown stubble beard, weathered tanned skin, and his thin gold wire round aviator-style glasses. NEVER use the bald man behind him. NEVER copy the black chef jacket, the kitchen, the pots, the food, the paper signs or the flash look.", "image_2": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "wardrobe": "A detective's working outfit, lived-in and slightly rumpled: a worn knee-length khaki-beige cotton gabardine trench coat with epaulettes, storm flap and a belt left hanging loose and unbuckled, coat worn open; underneath, a charcoal grey wool two-piece suit, single-breasted, the jacket unbuttoned; a pale blue cotton dress shirt with the top button undone; a dark navy knit tie loosened and pulled slightly off-centre; brown leather belt; scuffed dark brown leather derby shoes; a thin leather-strap wristwatch; his thin gold wire round glasses from Image 1. In his left hand a small worn black notebook with an elastic band and a short pencil. No badge, no gun, no logos, no printed text on any clothing.", "layout": { "top_row": [ "Full-body front, weight on one leg, right hand in trouser pocket pushing the coat back, notebook in the left hand, tired skeptical look, eyes to lens", "Full-body left profile, trench coat silhouette with the loose belt hanging, slight forward lean of a man who has been standing all day", "Full-body back view, back of the trench coat with its yoke and vent, belt ends hanging, grey-flecked hair at the nape", "Large head-and-shoulders close-up, front, glasses on, a dry half-squint of suspicion, loosened tie and open collar, coat collar half turned up", "Three-quarter close-up turned to his left, looking up from the open notebook toward the camera over the top of his glasses, pencil in hand" ], "bottom_row_labels_exact": ["GREY-FLECKED HAIR", "TRENCH COAT", "LOOSE TIE", "WIRE GLASSES", "NOTEBOOK", "SKIN TEXTURE / STUBBLE", "COLOR PALETTE"], "palette_swatches": ["trench khaki", "charcoal grey", "pale blue", "navy", "dark brown leather", "weathered skin tan"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the shoes, light grey seamless paper with a faint floor curve.", "realism": "Natural, real, slightly imperfect skin: visible pores, deep forehead lines, crow's feet, nasolabial folds, faint under-eye bags and darkness, mild sun damage and a few small age spots, slight redness on the nose and cheeks, uneven stubble density with grey patches, a few stray brow hairs, slightly dry lips, a faint red mark on the nose bridge from the glasses. No retouching. Lived-in clothes: creases at the elbows and behind the knees, slightly frayed trench cuffs, a faint stain near the coat hem, worn shoe toes, dog-eared notebook pages. Use case: professional casting and wardrobe board.", "constraints": "Same man in every panel, his face comes only from the foreground man in Image 1. Glasses on in every panel. No chef jacket, no kitchen, no second man, no hat, no beauty retouching, no plastic skin, no perfect symmetry, no watermark, no text other than the seven labels." }

Prompt 5

GPT IMAGE 2.5 · CHARACTER CHART · THE WITNESS { "task": "Photorealistic character reference sheet for AI film production, assembled from real photographs taken in one studio session. The character is an ordinary elderly woman who has just been scammed and lost her savings; she is sad, tired and quietly heartbroken.", "references": { "image_1": "IDENTITY ONLY. This exact elderly woman: her face, bone structure, dark brown hooded eyes, deep forehead lines, crow's feet, nasolabial folds, soft sagging jawline, rosy patches on the cheeks, small dark spot on her left cheek, full pale lips, and her salt-and-pepper grey hair loosely pinned up in a messy bun with many loose wispy strands falling around the face and neck, held by a small bronze hairpin. NEVER copy the dark grey jacket, the red patterned scarf, the white backdrop or the watermark.", "image_2": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "wardrobe": "Ordinary everyday clothes that clearly were not chosen with care, put on in a hurry and not matching each other: a faded lilac acrylic cardigan with pilling, buttoned one button off so it hangs crooked; underneath, a faded mustard-and-brown small floral print polyester blouse with the collar folded wrong on one side; loose shapeless brown polyester trousers slightly too short, with a faint old stain on one knee; thick beige ribbed socks; worn flat dark-blue velcro walking shoes with creased uppers; a cheap old black faux-leather handbag with peeling edges clutched against her body; in her right hand an old cracked smartphone with a plain case. No logos, no printed text on any clothing.", "expression": "Genuinely sad and defeated: eyes slightly red-rimmed and wet, brows drawn up in the middle, mouth pressed and pulled down at the corners, shoulders slumped and rounded forward, the look of someone who has just realised she was deceived. Never theatrical crying, never comic.", "layout": { "top_row": [ "Full-body front, slumped posture, both hands holding the handbag in front of her stomach with the phone in one hand, looking down and slightly away from the lens", "Full-body left profile, rounded back and forward-hanging head, cardigan hanging crooked, handbag pressed to her side", "Full-body back view, tired rounded shoulders, messy grey bun with the bronze hairpin, crooked cardigan hem", "Large head-and-shoulders close-up, front, wet red-rimmed eyes looking into the lens with quiet despair, loose grey strands across the cheek", "Three-quarter close-up turned to her right, staring down at the cracked phone screen held near her chest, the other hand pressed against her lips" ], "bottom_row_labels_exact": ["GREY BUN", "CROOKED CARDIGAN", "FLORAL BLOUSE", "OLD HANDBAG", "CRACKED PHONE", "SKIN TEXTURE", "COLOR PALETTE"], "palette_swatches": ["faded lilac", "mustard brown", "dull brown", "black", "salt-and-pepper grey", "warm aged skin"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the shoes, light grey seamless paper with a faint floor curve. Slightly flat, unflattering light.", "realism": "Unretouched aged skin: deep wrinkles, visible pores, fine crepey texture, age spots, uneven redness, slight puffiness under the eyes, dry lips. Real fabric: acrylic pilling, washed-out colours, uneven seams, creased polyester, scuffed shoes, peeling handbag edges, a spiderweb crack across the phone screen. Use case: professional casting and wardrobe board.", "constraints": "Same woman in every panel, her face comes only from Image 1. Same sad expression throughout. No jewellery other than the bronze hairpin, no beauty retouching, no plastic skin, no extra people, no watermark, no text other than the seven labels." }

Prompt 6

GPT IMAGE 2.5 SUNBURST · LINEUP STILL · 16:9 { "task": "A single photorealistic film still for an AI short film. Real photograph, not a composite.", "references": { "image_1": "LINEUP ROOM ONLY. Copy the police lineup room: the white wall with black horizontal height lines and red-and-white height markings from 2'6\" to 7'0\" on both sides, the hanging fluorescent tube lights, the dark band along the bottom of the wall, the grey floor, the cold flat light. NEVER copy the five people standing in it or the camera in the left foreground.", "image_2": "TWO CHARACTER SHEETS. Top sheet = SUSPECT 1, the young woman with the platinum ponytail, burgundy chunky-knit sweater, cream scarf, denim mini skirt, black tights, biker boots, silver hoops and nose stud. Bottom sheet = SUSPECT 2, the young man with messy black middle-part hair, black rectangular glasses, cream waffle henley, navy pinstripe wide trousers, suede sneakers, pearl studs, wired earphones. Take only faces, hair, bodies and outfits. Ignore grids, labels and grey backgrounds.", "image_3": "TWO CHARACTER SHEETS. Top sheet = SUSPECT 3, the heavyset man with the waxed handlebar moustache, wavy shoulder-length hair, teal cap with black headphones over it, round glasses, white psychedelic graphic tee, olive shorts, white socks, chunky beige sneakers. Bottom sheet = SUSPECT 4, the man with curly dark brown hair, moustache and goatee, glossy burgundy leather bomber jacket, white shirt, cream jeans, burgundy running sneakers, no sunglasses. Take only faces, hair, bodies and outfits. Ignore grids, labels and grey backgrounds.", "image_4": "CHARACTER SHEET, THE ELDERLY WOMAN. Grey messy bun with a bronze hairpin, crooked faded lilac cardigan, floral blouse, brown trousers, old black handbag. Take only her, ignore the grid and labels.", "image_5": "CHARACTER SHEET, THE DETECTIVE. Grey-flecked short hair, stubble, thin gold wire glasses, open khaki trench coat over a charcoal suit, loosened navy tie, small black notebook. Take only him, ignore the grid and labels." }, "scene": "A dim police observation room looking through a large one-way glass window into the brightly lit lineup room next door. The window fills most of the frame, set in a thin brushed-steel frame on a dark grey wall with a narrow ledge below it. The observation room is dark, lit only by the spill of cold light coming through the glass.", "blocking": "In the foreground, seen from behind at a rear diagonal over their shoulders, the elderly woman from Image 4 on the left and the detective from Image 5 on the right stand close to the glass looking in. We see the back of her grey bun with the bronze hairpin, her rounded shoulders in the lilac cardigan, her hand clutching the old handbag strap; next to her the detective's shoulder and back in the khaki trench coat, his head turned slightly so part of his stubbled jaw and the arm of his glasses show in profile, notebook in his hand at his side. Both are dark silhouettes rimmed by the cold light from the glass, slightly soft in focus. Through the glass, sharp and evenly lit, four suspects stand side by side against the height chart wall, facing the glass, each holding a white card at chest height with a single large black number: from left to right, SUSPECT 1 the woman from Image 2 holding '1', SUSPECT 2 the young man from Image 2 holding '2', SUSPECT 3 the moustached man from Image 3 holding '3', SUSPECT 4 the curly-haired man from Image 3 holding '4'. Each stands upright with a different attitude: the woman bored and cold, the young man nervous, the moustached man deadpan, the curly-haired man relaxed and unbothered.", "glass": "Faint real one-way-glass reflections: a ghostly reflection of the old woman's and the detective's faces barely visible on the glass, a soft sheen and a few smudges, without hiding the suspects.", "camera": "A real photograph taken on an iPhone with the texture of an ARRI Alexa and ARRI Signature Prime lenses, focus on the suspects behind the glass, the two foreground figures soft, gentle low-contrast rendering, lifted blacks, soft highlight roll-off, muted desaturated grade, fine grain.", "light": "Lineup room: cold, flat, slightly green-white fluorescent light from the hanging tubes, hard enough to show every face clearly. Observation room: almost dark, the two foreground figures lit only by the cold rim light through the glass.", "realism": "Unretouched skin with natural pores on every face, real fabric texture, scuffs on the lineup floor, slightly uneven paint on the height wall. Each suspect keeps the exact face from their character sheet.", "constraints": "The cards read only 1, 2, 3, 4 in large black numerals. Height markings stay as on Image 1. No extra people, no captions, no labels from the character sheets, no watermark." }

Prompt 7

GPT IMAGE 2.5 SUNBURST · BOOK · VOL. 02 · 9:16 Create a new product sheet for VOL. 02 of the same book, based on Image 1. Keep from Image 1 exactly: the four-view layout on a clean white background (front cover top-left, standing angled view with spine top-right, back cover bottom-left, book lying flat with a red ribbon bookmark bottom-right), the black linen hardcover with its fine weave texture, the soft studio shadows, and all typography in the same fonts, sizes and positions: "FIELD MANUAL", the title "ULTIMATE CINEMATIC REALISM BLUEPRINT" in large white condensed capitals, the arcads logo under the title, "A FIELD MANUAL FOR PHOTOREAL AI FILMMAKING" along the bottom, the same spine title, and on the back cover "EVERY FRAME IS A DECISION.", the arcads logo bottom-left and the barcode bottom-right. Change only these two things: 1. "VOL. 01" becomes "VOL. 02" on the front cover, the spine and the lying-flat view. 2. The black film strip with sprocket holes now holds SIX portrait frames instead of four, numbered 01 02 03 04 05 06 underneath, on every view where the strip appears. Each frame is a tight head-and-shoulders portrait on a soft grey studio background with the same muted cinematic grade: Frame 01: the young woman from the top sheet of Image 2, sleek platinum ponytail, silver hoop earrings and tiny nose stud, burgundy sweater and cream scarf, cool stare. Frame 02: the young man from the bottom sheet of Image 2, messy black middle-part hair, black rectangular glasses, pearl stud, cream henley collar, wired earphone. Frame 03: the heavyset man from the top sheet of Image 3, teal cap with black headphones over it, round glasses, waxed handlebar moustache, white tee collar. Frame 04: the curly-haired man from the bottom sheet of Image 3, curly dark brown hair, moustache and goatee, burgundy leather jacket collar over a white shirt, no sunglasses. Frame 05: the elderly woman from Image 4, grey messy bun with the bronze hairpin, sad tired eyes, lilac cardigan and floral blouse collar. Frame 06: the detective from Image 5, grey-flecked hair, stubble, thin gold wire glasses, khaki trench coat collar and loosened navy tie, a skeptical look. From Images 2 to 5 take only faces, hair and the visible top of each outfit; ignore their grids, labels and full-body panels. Every face must match its character sheet exactly. The six frames fit evenly across the cover width inside the same strip, printed on the cover with the correct perspective on the angled and lying-flat views. No other changes, no extra text.

Prompt 8

GPT IMAGE 2.5 SUNBURST · BOOK EDIT · BELLY BAND Edit Image 1. Keep everything exactly the same: the four views of the black linen hardcover (front cover, angled view with spine, back cover, lying flat with red ribbon), the white background, the shadows, all existing typography, the arcads logos, the film strip with the six portraits and the numbers 01 to 06, the barcode. Add only one thing: a printed paper belly band (an obi band) wrapped horizontally around the lower part of the book, covering only the area below the film strip where "A FIELD MANUAL FOR PHOTOREAL AI FILMMAKING" sits, so the title, the arcads logo and all six portraits stay fully visible. The band is a strip of matte paper in the deep red of the ribbon bookmark, wrapping around the front, the spine and the back cover, with clean white condensed type centred on the front: "FREE TRIAL · 1,000 FREE CREDITS" and a smaller line beneath it: "FOR ALL NEW ARCADS CUSTOMERS" On the spine part of the band, only a small white arcads mark. The band shows a real paper edge, a slight gap where it lifts off the cover and soft shadow, following the perspective on the angled and lying-flat views; on the back cover view the band continues plain red with no text. Spell every word exactly as written. No other change anywhere.

Tutorial

DENIZ AKKABAK · REAL OR NOT TUTORIAL · @DENIZ.AKKABAK

MADE IN ARCADS · arcads.ai

The finished 30-second film — one witness, one detective, four suspects, one book. (Open this document in Google Docs to watch it play.)
The finished 30-second film — one witness, one detective, four suspects, one book. (Open this document in Google Docs to watch it play.)

ULTIMATE CINEMATIC REALISM BLUEPRINT · VOL. 02

Real or Not

Lock six characters into studio charts, stage a police lineup behind one-way glass, print a hero book with a live offer on its belly band, and direct a 30-second lineup film with six voices — every step made in Arcads.

A TUTORIAL BY DENIZ AKKABAK · @DENIZ.AKKABAK

PIPELINE GPT IMAGE 2.5 → GPT IMAGE 2.5 SUNBURST → SEEDANCE 2.5 FORMAT 4:3 · 30 S · 1080P PLATFORM ARCADS

The first film asked a bouncer to spot the AI. This one hands the job to the most honest judge there is: a tired old woman behind one-way glass, with a detective at her shoulder asking the only question that matters — which ones are AI? Four suspects hold their numbers. One chews gum like nothing can touch her, one sweats through his cool, one cannot stop saying it was not him, and one just smiles. The joke works only if the audience plays witness too and genuinely cannot tell who is synthetic, which makes realism the punchline again. This blueprint walks through the whole build exactly as we made it in Arcads: six character charts, one multi-character lineup still, a VOL. 02 hero book with a printed offer band, and a single 30-second Seedance 2.5 generation with nine scoped references. Every prompt is printed verbatim, and every “why it works” points at the exact line that does the work.

THE PIPELINE AT A GLANCE

  1. 01

    Cast the six

    One face and one outfit per suspect; the witness and the detective get a face and a written wardrobe. The face owns identity, the outfit owns clothing, and neither is allowed to borrow from the other.

  2. 02

    Lock the charts

    GPT Image 2.5 turns every pair into a five-angle studio sheet with seven labelled detail tiles — the continuity bible every later model restores from.

  3. 03

    Stage the lineup

    GPT Image 2.5 Sunburst builds the observation room, the one-way glass and all six people in one still. Two charts per reference image get six characters past a five-image limit.

  4. 04

    Print the prop

    The VOL. 01 product sheet becomes VOL. 02 with six portraits, then a single edit wraps it in a red belly band carrying the Arcads free-trial offer.

  5. 05

    Shoot the film

    Seedance 2.5 reads nine references and directs a continuous lineup track, six voices, three arrests and a final book reveal in one 30-second generation.

THE MODEL MAP — WHICH MODEL FOR WHICH JOB

  1. 01

    GPT Image 2.5

    Every character chart. It obeys reference scopes literally and spells labels cleanly. Charts ran at 16:9 with the same three-slot JSON architecture.

  2. 02

    GPT Image 2.5 Sunburst

    The multi-character still and both book edits. For the lineup we ran the same prompt through GPT Image 2.5, Sunburst and Nano Banana 2.1 side by side and kept the Sunburst frame.

  3. 03

    Seedance 2.5

    The director. Reference to Video, 30 seconds, multi-shot editing, native dialogue and sound: 4:3, 1080p, audio on, high bitrate, two variants of the same prompt to choose from.

STEP 01 — PINTEREST → ARCADS

Casting the Lineup

Realism starts before the first prompt. Each suspect is two pictures: a face, which owns bone structure, skin and hair, and an outfit, which owns cut, fabric and colour. Keeping them in separate images is the entire trick — a model can only blend what you hand it inside one frame, so the outfit photo’s own face is explicitly banned in the prompt. The witness and the detective needed no outfit photo at all: a scam victim’s mismatched cardigan and a rumpled trench coat are easier to write than to find. For suspect four I skipped casting and used my own chart.

REFERENCE INPUTS — ONE FACE, ONE OUTFIT

Face — suspect 1
Face — suspect 1
Outfit — burgundy knit, cream scarf, denim mini
Outfit — burgundy knit, cream scarf, denim mini
Face — suspect 2
Face — suspect 2
Outfit — waffle henley, pinstripe wide trousers
Outfit — waffle henley, pinstripe wide trousers
Face — suspect 3
Face — suspect 3
Outfit — graphic tee, olive shorts, cap and headphones
Outfit — graphic tee, olive shorts, cap and headphones
Face — the witness; her wardrobe is written
Face — the witness; her wardrobe is written
Face — the detective (foreground man only); wardrobe written
Face — the detective (foreground man only); wardrobe written

STEP 02 — GPT IMAGE 2.5

Locking the Cast — Character Charts

Every chart uses the same architecture: an IDENTITY ONLY slot, a WARDROBE ONLY slot and a LAYOUT ONLY slot, each closed with a NEVER clause that names what it must not contribute. The layout slot is simply a finished chart from the previous film — it teaches the grid, the grey seamless and the label typography without lending a face. Where a character has no outfit photo, the wardrobe lives in its own written block, and where the story needs an emotion, it gets an expression block of its own.

THE LAYOUT TEMPLATE — ATTACHED LAST IN EVERY CHART

A finished chart from the first film. It contributes structure only; its woman, outfit and labels are banned in every prompt.
A finished chart from the first film. It contributes structure only; its woman, outfit and labels are banned in every prompt.

Chart 1 — suspect 1, plus new jewellery

The jewellery block adds what neither reference has — hoops, a helix stud and a tiny nose stud — and pins it to every panel where an ear or the face shows.

REFERENCE INPUTS — ATTACH IN THIS ORDER

Image 1 — the face. Image 2 — the outfit. Image 3 — the layout template.

Prompt 1

GPT IMAGE 2.5 · CHARACTER CHART · SUSPECT 1 { "task": "Photorealistic fashion character reference sheet for AI film production, assembled from real photographs taken in one studio session.", "references": { "image_1": "IDENTITY ONLY. This exact young woman: her face, bone structure, pale blue-grey eyes, thick dark brushed-up brows, straight nose, full soft pink lips, fair skin with natural freckles across the nose, cheeks and shoulders, and her platinum ash-blonde hair pulled back tight into a high sleek ponytail with a few loose flyaway strands, the small dark fine-line tattoo behind her ear on the side of the neck. NEVER copy the bare shoulder or the shoulder tattoo exposure, the dark background or the harsh hard sunlight.", "image_2": "WARDROBE ONLY. Copy the outfit exactly: oversized chunky-knit burgundy crewneck sweater with ribbed cuffs and hem, a long cream chunky-knit scarf draped around the neck with both ends hanging down the front, a light-wash blue denim mini skirt, sheer black tights, white ribbed knee socks folded over the top of black leather knee-high biker boots with buckle straps, and a black slouchy leather shoulder bag with silver hardware. NEVER copy her face, her long dark curly hair, the phone, the mirror, the bedroom or the selfie pose.", "image_3": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "jewellery": "New additions not in either reference: small polished silver hoop earrings in both earlobes plus a tiny silver stud in the upper helix of her left ear, and a delicate, elegant tiny silver stud piercing on the left side of her nose, minimal and chic. Visible in every panel where her face or ears show.", "layout": { "top_row": [ "Full-body front, relaxed stance, one hand at her side, bag on her left shoulder, cool neutral expression, eyes to lens", "Full-body left profile, chin slightly raised, ponytail clearly visible, scarf ends hanging forward, boots and folded socks in clean silhouette", "Full-body back view, sleek platinum ponytail, back of the burgundy sweater, scarf around the neck, bag strap over the shoulder", "Large head-and-shoulders close-up, front, looking straight into the lens, freckles, hoop earrings and nose stud sharp, scarf around the neck", "Three-quarter close-up turned to her right, looking back over her shoulder at the camera like Image 1, ear with hoop and helix stud and the small neck tattoo visible, nose stud catching a soft highlight" ], "bottom_row_labels_exact": ["PLATINUM PONYTAIL", "SILVER JEWELLERY", "CHUNKY KNIT", "CREAM SCARF", "DENIM SKIRT", "BIKER BOOTS", "COLOR PALETTE"], "palette_swatches": ["burgundy", "cream", "light denim blue", "black", "platinum blonde", "silver"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the boots, light grey seamless paper with a faint floor curve.", "realism": "Unretouched skin with visible pores, real freckles and fine peach fuzz, natural lip texture, real flyaway hairs at the hairline. Knit texture with visible stitches and a little pilling, denim weave and fading, matte sheerness of the tights showing skin tone, creased leather on the boots and bag, small reflections on the silver jewellery. Use case: professional casting and wardrobe board.", "constraints": "Same woman in every panel, her face comes only from Image 1. Platinum ponytail in every panel, never long dark curls. Earrings and nose stud consistent in every panel. No phone, no mirror, no bedroom. No beauty retouching, no plastic skin, no extra people, no watermark, no text other than the seven labels." }

OUTPUT

GPT Image 2.5 · 16:9 — Platinum ponytail from the face, burgundy knit from the outfit, silver added by text.
GPT Image 2.5 · 16:9 — Platinum ponytail from the face, burgundy knit from the outfit, silver added by text.

Chart 2 — suspect 2

His glasses come from the outfit photo, so the prompt files them under wardrobe and then locks them on in every panel — an accessory that defines a character must never be optional.

REFERENCE INPUTS — ATTACH IN THIS ORDER

Image 1 — the face. Image 2 — the outfit. Image 3 — the layout template.

Prompt 2

GPT IMAGE 2.5 · CHARACTER CHART · SUSPECT 2 { "task": "Photorealistic fashion character reference sheet for AI film production, assembled from real photographs taken in one studio session.", "references": { "image_1": "IDENTITY ONLY. This exact young man: his face, bone structure, sharp jawline, dark brown almond eyes, straight dark brows, straight nose, full lips, light stubble on the upper lip and chin, warm light skin with visible pores, his messy black middle-parted hair falling in loose strands over the forehead and around the eyes, and the small round pearl stud earrings in both ears. NEVER copy the bare chest, the white backdrop or the wet skin sheen.", "image_2": "WARDROBE ONLY. Copy the outfit exactly: an off-white cream waffle-knit long-sleeve henley with a three-button placket, worn untucked and loose, sleeves pushed up slightly on the forearms; navy pinstripe wide-leg pleated trousers with a high rise and a long full break; cream and grey suede low-top sneakers with gum-coloured soles; black rectangular acetate glasses; white wired earphones hanging from his ears; a thin silver chain necklace; a silver metal-strap wristwatch on his right wrist; a couple of thin silver rings; a black nylon crossbody shoulder bag on his left shoulder. NEVER copy his face, his wavy dark hair, the parking lot or the street.", "image_3": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "layout": { "top_row": [ "Full-body front, relaxed stance, left hand in trouser pocket, bag strap over the shoulder, calm neutral expression, glasses on, eyes to lens", "Full-body left profile, mid-stride walking pose, the wide trouser legs swinging, earphone wire visible", "Full-body back view, back of the henley, bag hanging on his left side, trouser hems pooling over the sneakers", "Large head-and-shoulders close-up, front, glasses on, loose black strands over the forehead, pearl stud and earphone visible, henley placket and silver chain sharp", "Three-quarter close-up turned slightly to his left, glasses on, looking just past the camera, profile of the jaw, pearl earring and earphone wire catching the light" ], "bottom_row_labels_exact": ["MIDDLE-PART HAIR", "WAFFLE HENLEY", "PINSTRIPE TROUSERS", "SUEDE SNEAKERS", "GLASSES & PEARLS", "SKIN TEXTURE", "COLOR PALETTE"], "palette_swatches": ["cream", "navy pinstripe", "black", "silver", "gum brown", "warm skin tone"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the sneakers, light grey seamless paper with a faint floor curve.", "realism": "Unretouched skin with visible pores, faint stubble texture, natural lip texture, real stray hairs. Waffle knit with visible grid texture and slight stretch at the elbows, fine pinstripes that stay straight across pleats and folds, suede nap on the sneakers, small reflections on the lenses, pearls and silver. Use case: professional casting and wardrobe board.", "constraints": "Same man in every panel, his face and hair come only from Image 1. Glasses on in every panel. No phone, no street, no extra people, no beauty retouching, no plastic skin, no logos, no watermark, no text other than the seven labels." }

OUTPUT

GPT Image 2.5 · 16:9 — Pearl studs from the face, glasses and earphones from the outfit, one identity.
GPT Image 2.5 · 16:9 — Pearl studs from the face, glasses and earphones from the outfit, one identity.

Chart 3 — suspect 3

Two details matter here. The tee graphic is rebuilt as an original illustration with no readable text, and a separate hair_with_cap line tells the model where his long hair goes once a cap and headphones sit on top of it.

REFERENCE INPUTS — ATTACH IN THIS ORDER

Image 1 — the face. Image 2 — the outfit. Image 3 — the layout template.

Prompt 3

GPT IMAGE 2.5 · CHARACTER CHART · SUSPECT 3 { "task": "Photorealistic fashion character reference sheet for AI film production, assembled from real photographs taken in one studio session.", "references": { "image_1": "IDENTITY ONLY. This exact man: heavyset build, broad round face with full cheeks and a soft double chin, hazel-brown eyes under thick dark brows, a rounded nose, a thick dark brown waxed handlebar moustache with curled tips, light stubble on the chin and jaw, fair skin with a ruddy flush on the cheeks and nose and a few freckles, and his shoulder-length messy wavy dark brown hair. NEVER copy the charcoal T-shirt or the white wall.", "image_2": "WARDROBE ONLY. Copy the outfit exactly: a teal-green baseball cap with a small red embroidered patch on the front, worn forward; black over-ear wireless headphones worn over the cap; thin round black wire glasses; an oversized boxy white cotton T-shirt with dropped shoulders and a colourful psychedelic illustrated graphic on the chest (cartoon flowers, small creatures and a pink figure, in bright pinks, blues, greens and yellows); dark olive textured ripstop shorts above the knee with a small yellow woven label tab on the left pocket; a black smartwatch; white ribbed crew socks pulled up; chunky beige-cream caged hiking-style sneakers with thick lugged soles. Keep the T-shirt graphic as an original illustration with no readable text. NEVER copy his face, the phone, the mirror, the hallway or the watermark.", "image_3": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "hair_with_cap": "His wavy dark brown hair spills out from under the cap and the headphones at the sides and back, falling to the shoulders.", "layout": { "top_row": [ "Full-body front, feet apart, one hand in his shorts pocket, cap, headphones and glasses on, deadpan expression, eyes to lens", "Full-body left profile, belly in silhouette, hair falling behind the headphone band, chunky sneakers in clean side view", "Full-body back view, back of the cap and headphone band, wavy hair over the collar, plain back of the white tee, olive shorts", "Large head-and-shoulders close-up, front, cap, headphones and round glasses on, handlebar moustache sharp, the top of the T-shirt graphic visible", "Three-quarter close-up turned to his right, adjusting the headphone ear cup with one hand, moustache curl and cap patch catching the light" ], "bottom_row_labels_exact": ["CAP & HEADPHONES", "HANDLEBAR MOUSTACHE", "GRAPHIC TEE", "OLIVE SHORTS", "CHUNKY SNEAKERS", "SKIN TEXTURE", "COLOR PALETTE"], "palette_swatches": ["teal green", "white", "olive drab", "cream beige", "black", "hot pink"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the sneakers, light grey seamless paper with a faint floor curve.", "realism": "Unretouched skin with visible pores, ruddy cheeks, slight oily shine on the nose and forehead, individual moustache hairs and wax sheen on the curled tips, frizzy flyaway strands in the wavy hair. Real body weight: the tee draping over the belly, natural creases at the sides. Cotton tee with slightly faded print texture, ripstop grid on the shorts, mesh and rubber on the sneakers, soft leatherette on the headphone cushions, small reflections on the glasses. Use case: professional casting and wardrobe board.", "constraints": "Same man in every panel, his face, hair and moustache come only from Image 1. Cap, headphones and glasses on in every panel. No phone, no mirror, no extra people, no beauty retouching, no plastic skin, no readable logos, no watermark, no text other than the seven labels." }

OUTPUT

GPT Image 2.5 · 16:9 — The handlebar moustache and the hair survive the cap and headphones.
GPT Image 2.5 · 16:9 — The handlebar moustache and the hair survive the cap and headphones.

Chart 4 — the detective

The face photo contains a second man, so the identity slot names the foreground man by pose and bans the other one outright. The wardrobe is written as lived-in rather than costume: a loose unbuckled belt, a frayed cuff, a faint stain near the hem.

REFERENCE INPUTS — ATTACH IN THIS ORDER

Image 1 — the face (a photo with two people). Image 2 — the layout template.

Prompt 4

GPT IMAGE 2.5 · CHARACTER CHART · THE DETECTIVE { "task": "Photorealistic character reference sheet for AI film production, assembled from real photographs taken in one studio session. The character is a seasoned city detective.", "references": { "image_1": "IDENTITY ONLY, THE MAN IN THE FOREGROUND. Take only the man in front with his arms crossed: his lean angular face, high cheekbones, deep-set light eyes, straight nose, thin lips, short grey-flecked dark brown hair, short grey and brown stubble beard, weathered tanned skin, and his thin gold wire round aviator-style glasses. NEVER use the bald man behind him. NEVER copy the black chef jacket, the kitchen, the pots, the food, the paper signs or the flash look.", "image_2": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "wardrobe": "A detective's working outfit, lived-in and slightly rumpled: a worn knee-length khaki-beige cotton gabardine trench coat with epaulettes, storm flap and a belt left hanging loose and unbuckled, coat worn open; underneath, a charcoal grey wool two-piece suit, single-breasted, the jacket unbuttoned; a pale blue cotton dress shirt with the top button undone; a dark navy knit tie loosened and pulled slightly off-centre; brown leather belt; scuffed dark brown leather derby shoes; a thin leather-strap wristwatch; his thin gold wire round glasses from Image 1. In his left hand a small worn black notebook with an elastic band and a short pencil. No badge, no gun, no logos, no printed text on any clothing.", "layout": { "top_row": [ "Full-body front, weight on one leg, right hand in trouser pocket pushing the coat back, notebook in the left hand, tired skeptical look, eyes to lens", "Full-body left profile, trench coat silhouette with the loose belt hanging, slight forward lean of a man who has been standing all day", "Full-body back view, back of the trench coat with its yoke and vent, belt ends hanging, grey-flecked hair at the nape", "Large head-and-shoulders close-up, front, glasses on, a dry half-squint of suspicion, loosened tie and open collar, coat collar half turned up", "Three-quarter close-up turned to his left, looking up from the open notebook toward the camera over the top of his glasses, pencil in hand" ], "bottom_row_labels_exact": ["GREY-FLECKED HAIR", "TRENCH COAT", "LOOSE TIE", "WIRE GLASSES", "NOTEBOOK", "SKIN TEXTURE / STUBBLE", "COLOR PALETTE"], "palette_swatches": ["trench khaki", "charcoal grey", "pale blue", "navy", "dark brown leather", "weathered skin tan"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the shoes, light grey seamless paper with a faint floor curve.", "realism": "Natural, real, slightly imperfect skin: visible pores, deep forehead lines, crow's feet, nasolabial folds, faint under-eye bags and darkness, mild sun damage and a few small age spots, slight redness on the nose and cheeks, uneven stubble density with grey patches, a few stray brow hairs, slightly dry lips, a faint red mark on the nose bridge from the glasses. No retouching. Lived-in clothes: creases at the elbows and behind the knees, slightly frayed trench cuffs, a faint stain near the coat hem, worn shoe toes, dog-eared notebook pages. Use case: professional casting and wardrobe board.", "constraints": "Same man in every panel, his face comes only from the foreground man in Image 1. Glasses on in every panel. No chef jacket, no kitchen, no second man, no hat, no beauty retouching, no plastic skin, no perfect symmetry, no watermark, no text other than the seven labels." }

OUTPUT

GPT Image 2.5 · 16:9 — Weathered, imperfect skin and a rumpled trench — a man who has been standing all day.
GPT Image 2.5 · 16:9 — Weathered, imperfect skin and a rumpled trench — a man who has been standing all day.

Chart 5 — the witness

Her story is in the clothes: a cardigan buttoned one button off, a collar folded wrong, trousers slightly too short. The expression block asks for real sadness and explicitly forbids theatrical crying and comedy, and the light is kept deliberately flat and unflattering.

REFERENCE INPUTS — ATTACH IN THIS ORDER

Image 1 — the face. Image 2 — the layout template.

Prompt 5

GPT IMAGE 2.5 · CHARACTER CHART · THE WITNESS { "task": "Photorealistic character reference sheet for AI film production, assembled from real photographs taken in one studio session. The character is an ordinary elderly woman who has just been scammed and lost her savings; she is sad, tired and quietly heartbroken.", "references": { "image_1": "IDENTITY ONLY. This exact elderly woman: her face, bone structure, dark brown hooded eyes, deep forehead lines, crow's feet, nasolabial folds, soft sagging jawline, rosy patches on the cheeks, small dark spot on her left cheek, full pale lips, and her salt-and-pepper grey hair loosely pinned up in a messy bun with many loose wispy strands falling around the face and neck, held by a small bronze hairpin. NEVER copy the dark grey jacket, the red patterned scarf, the white backdrop or the watermark.", "image_2": "LAYOUT ONLY. Copy the character sheet structure: five top panels (full-body front, full-body side profile, full-body back, large head-and-shoulders close-up, three-quarter close-up), seven square detail tiles along the bottom with uppercase sans-serif labels above them, the light grey seamless background, the soft studio light and the panel proportions. NEVER copy that woman, her outfit, her colours or her label words." }, "wardrobe": "Ordinary everyday clothes that clearly were not chosen with care, put on in a hurry and not matching each other: a faded lilac acrylic cardigan with pilling, buttoned one button off so it hangs crooked; underneath, a faded mustard-and-brown small floral print polyester blouse with the collar folded wrong on one side; loose shapeless brown polyester trousers slightly too short, with a faint old stain on one knee; thick beige ribbed socks; worn flat dark-blue velcro walking shoes with creased uppers; a cheap old black faux-leather handbag with peeling edges clutched against her body; in her right hand an old cracked smartphone with a plain case. No logos, no printed text on any clothing.", "expression": "Genuinely sad and defeated: eyes slightly red-rimmed and wet, brows drawn up in the middle, mouth pressed and pulled down at the corners, shoulders slumped and rounded forward, the look of someone who has just realised she was deceived. Never theatrical crying, never comic.", "layout": { "top_row": [ "Full-body front, slumped posture, both hands holding the handbag in front of her stomach with the phone in one hand, looking down and slightly away from the lens", "Full-body left profile, rounded back and forward-hanging head, cardigan hanging crooked, handbag pressed to her side", "Full-body back view, tired rounded shoulders, messy grey bun with the bronze hairpin, crooked cardigan hem", "Large head-and-shoulders close-up, front, wet red-rimmed eyes looking into the lens with quiet despair, loose grey strands across the cheek", "Three-quarter close-up turned to her right, staring down at the cracked phone screen held near her chest, the other hand pressed against her lips" ], "bottom_row_labels_exact": ["GREY BUN", "CROOKED CARDIGAN", "FLORAL BLOUSE", "OLD HANDBAG", "CRACKED PHONE", "SKIN TEXTURE", "COLOR PALETTE"], "palette_swatches": ["faded lilac", "mustard brown", "dull brown", "black", "salt-and-pepper grey", "warm aged skin"] }, "camera": "Full-body panels at 85mm f/8, close-ups at 100mm f/5.6, eye-level, identical white balance across panels.", "lighting": "Large soft octabox key from front-left, gentle fill, soft contact shadows under the shoes, light grey seamless paper with a faint floor curve. Slightly flat, unflattering light.", "realism": "Unretouched aged skin: deep wrinkles, visible pores, fine crepey texture, age spots, uneven redness, slight puffiness under the eyes, dry lips. Real fabric: acrylic pilling, washed-out colours, uneven seams, creased polyester, scuffed shoes, peeling handbag edges, a spiderweb crack across the phone screen. Use case: professional casting and wardrobe board.", "constraints": "Same woman in every panel, her face comes only from Image 1. Same sad expression throughout. No jewellery other than the bronze hairpin, no beauty retouching, no plastic skin, no extra people, no watermark, no text other than the seven labels." }

OUTPUT

GPT Image 2.5 · 16:9 — A scam victim dressed in a hurry — the sadness is in the posture before it is in the face.
GPT Image 2.5 · 16:9 — A scam victim dressed in a hurry — the sadness is in the posture before it is in the face.

Chart 6 — suspect 4 (me)

I play suspect four, so I used my own chart, built earlier with the same system. In the film it carries the outfit, and a plain selfie joins it later as a face-support reference to sharpen the likeness.

My own chart — burgundy leather bomber, white shirt, cream denim.
My own chart — burgundy leather bomber, white shirt, cream denim.

STEP 03 — GPT IMAGE 2.5 SUNBURST

Staging the Lineup

The scene needs the room, four suspects, the witness and the detective — seven references for a model that accepts five. The fix is mechanical: stack two finished charts into one image, top and bottom, and name them inside the slot as “Top sheet = SUSPECT 1 … Bottom sheet = SUSPECT 2”. The lineup room itself comes from a single film still, scoped to the wall, the height markings and the fluorescent tubes, with its five actors banned. Then the prompt does the staging: the witness and the detective in the dark foreground as rim-lit silhouettes, the four suspects sharp behind the glass with their cards, and a ghost of their reflections on the glass so the window reads as a real window.

REFERENCE INPUTS — ATTACH IN THIS ORDER

Image 1 — the lineup room only
Image 1 — the lineup room only
Image 2 — suspects 1 and 2, stacked
Image 2 — suspects 1 and 2, stacked
Image 3 — suspects 3 and 4, stacked
Image 3 — suspects 3 and 4, stacked
Image 4 — the witness chart
Image 4 — the witness chart
Image 5 — the detective chart
Image 5 — the detective chart

Prompt 6

GPT IMAGE 2.5 SUNBURST · LINEUP STILL · 16:9 { "task": "A single photorealistic film still for an AI short film. Real photograph, not a composite.", "references": { "image_1": "LINEUP ROOM ONLY. Copy the police lineup room: the white wall with black horizontal height lines and red-and-white height markings from 2'6\" to 7'0\" on both sides, the hanging fluorescent tube lights, the dark band along the bottom of the wall, the grey floor, the cold flat light. NEVER copy the five people standing in it or the camera in the left foreground.", "image_2": "TWO CHARACTER SHEETS. Top sheet = SUSPECT 1, the young woman with the platinum ponytail, burgundy chunky-knit sweater, cream scarf, denim mini skirt, black tights, biker boots, silver hoops and nose stud. Bottom sheet = SUSPECT 2, the young man with messy black middle-part hair, black rectangular glasses, cream waffle henley, navy pinstripe wide trousers, suede sneakers, pearl studs, wired earphones. Take only faces, hair, bodies and outfits. Ignore grids, labels and grey backgrounds.", "image_3": "TWO CHARACTER SHEETS. Top sheet = SUSPECT 3, the heavyset man with the waxed handlebar moustache, wavy shoulder-length hair, teal cap with black headphones over it, round glasses, white psychedelic graphic tee, olive shorts, white socks, chunky beige sneakers. Bottom sheet = SUSPECT 4, the man with curly dark brown hair, moustache and goatee, glossy burgundy leather bomber jacket, white shirt, cream jeans, burgundy running sneakers, no sunglasses. Take only faces, hair, bodies and outfits. Ignore grids, labels and grey backgrounds.", "image_4": "CHARACTER SHEET, THE ELDERLY WOMAN. Grey messy bun with a bronze hairpin, crooked faded lilac cardigan, floral blouse, brown trousers, old black handbag. Take only her, ignore the grid and labels.", "image_5": "CHARACTER SHEET, THE DETECTIVE. Grey-flecked short hair, stubble, thin gold wire glasses, open khaki trench coat over a charcoal suit, loosened navy tie, small black notebook. Take only him, ignore the grid and labels." }, "scene": "A dim police observation room looking through a large one-way glass window into the brightly lit lineup room next door. The window fills most of the frame, set in a thin brushed-steel frame on a dark grey wall with a narrow ledge below it. The observation room is dark, lit only by the spill of cold light coming through the glass.", "blocking": "In the foreground, seen from behind at a rear diagonal over their shoulders, the elderly woman from Image 4 on the left and the detective from Image 5 on the right stand close to the glass looking in. We see the back of her grey bun with the bronze hairpin, her rounded shoulders in the lilac cardigan, her hand clutching the old handbag strap; next to her the detective's shoulder and back in the khaki trench coat, his head turned slightly so part of his stubbled jaw and the arm of his glasses show in profile, notebook in his hand at his side. Both are dark silhouettes rimmed by the cold light from the glass, slightly soft in focus. Through the glass, sharp and evenly lit, four suspects stand side by side against the height chart wall, facing the glass, each holding a white card at chest height with a single large black number: from left to right, SUSPECT 1 the woman from Image 2 holding '1', SUSPECT 2 the young man from Image 2 holding '2', SUSPECT 3 the moustached man from Image 3 holding '3', SUSPECT 4 the curly-haired man from Image 3 holding '4'. Each stands upright with a different attitude: the woman bored and cold, the young man nervous, the moustached man deadpan, the curly-haired man relaxed and unbothered.", "glass": "Faint real one-way-glass reflections: a ghostly reflection of the old woman's and the detective's faces barely visible on the glass, a soft sheen and a few smudges, without hiding the suspects.", "camera": "A real photograph taken on an iPhone with the texture of an ARRI Alexa and ARRI Signature Prime lenses, focus on the suspects behind the glass, the two foreground figures soft, gentle low-contrast rendering, lifted blacks, soft highlight roll-off, muted desaturated grade, fine grain.", "light": "Lineup room: cold, flat, slightly green-white fluorescent light from the hanging tubes, hard enough to show every face clearly. Observation room: almost dark, the two foreground figures lit only by the cold rim light through the glass.", "realism": "Unretouched skin with natural pores on every face, real fabric texture, scuffs on the lineup floor, slightly uneven paint on the height wall. Each suspect keeps the exact face from their character sheet.", "constraints": "The cards read only 1, 2, 3, 4 in large black numerals. Height markings stay as on Image 1. No extra people, no captions, no labels from the character sheets, no watermark." }

OUTPUT

GPT Image 2.5 Sunburst · 16:9 — the frame the film opens on, and the blocking map for everything after it.
GPT Image 2.5 Sunburst · 16:9 — the frame the film opens on, and the blocking map for everything after it.

STEP 04 — GPT IMAGE 2.5 SUNBURST

Printing the Prop — VOL. 02

The ending pays off with an object, so the object is designed before the film. The VOL. 01 product sheet from the first film is the base: same black linen, same title, same arcads logo, same four views. The prompt changes exactly two things — the volume number and a film strip that now holds six portraits drawn from the charts — and states everything else as a keep list.

REFERENCE INPUTS — ATTACH IN THIS ORDER

Image 1 — the VOL. 01 product sheet. Image 2 — suspects 1 and 2, stacked. Image 3 — suspects 3 and 4, stacked. Image 4 — the witness chart. Image 5 — the detective chart.

Prompt 7

GPT IMAGE 2.5 SUNBURST · BOOK · VOL. 02 · 9:16 Create a new product sheet for VOL. 02 of the same book, based on Image 1. Keep from Image 1 exactly: the four-view layout on a clean white background (front cover top-left, standing angled view with spine top-right, back cover bottom-left, book lying flat with a red ribbon bookmark bottom-right), the black linen hardcover with its fine weave texture, the soft studio shadows, and all typography in the same fonts, sizes and positions: "FIELD MANUAL", the title "ULTIMATE CINEMATIC REALISM BLUEPRINT" in large white condensed capitals, the arcads logo under the title, "A FIELD MANUAL FOR PHOTOREAL AI FILMMAKING" along the bottom, the same spine title, and on the back cover "EVERY FRAME IS A DECISION.", the arcads logo bottom-left and the barcode bottom-right. Change only these two things: 1. "VOL. 01" becomes "VOL. 02" on the front cover, the spine and the lying-flat view. 2. The black film strip with sprocket holes now holds SIX portrait frames instead of four, numbered 01 02 03 04 05 06 underneath, on every view where the strip appears. Each frame is a tight head-and-shoulders portrait on a soft grey studio background with the same muted cinematic grade: Frame 01: the young woman from the top sheet of Image 2, sleek platinum ponytail, silver hoop earrings and tiny nose stud, burgundy sweater and cream scarf, cool stare. Frame 02: the young man from the bottom sheet of Image 2, messy black middle-part hair, black rectangular glasses, pearl stud, cream henley collar, wired earphone. Frame 03: the heavyset man from the top sheet of Image 3, teal cap with black headphones over it, round glasses, waxed handlebar moustache, white tee collar. Frame 04: the curly-haired man from the bottom sheet of Image 3, curly dark brown hair, moustache and goatee, burgundy leather jacket collar over a white shirt, no sunglasses. Frame 05: the elderly woman from Image 4, grey messy bun with the bronze hairpin, sad tired eyes, lilac cardigan and floral blouse collar. Frame 06: the detective from Image 5, grey-flecked hair, stubble, thin gold wire glasses, khaki trench coat collar and loosened navy tie, a skeptical look. From Images 2 to 5 take only faces, hair and the visible top of each outfit; ignore their grids, labels and full-body panels. Every face must match its character sheet exactly. The six frames fit evenly across the cover width inside the same strip, printed on the cover with the correct perspective on the angled and lying-flat views. No other changes, no extra text.

Input — VOL. 01
Input — VOL. 01
Output — VOL. 02, six portraits
Output — VOL. 02, six portraits

The offer band

Arcads currently gives new customers a free trial with 1,000 free credits, and the book is the one object the camera is guaranteed to hold on. We tested three placements as single edits on the VOL. 02 sheet — a round sticker, a printed line under the strip, and a paper belly band — and kept the band: it reads in a 1.5-second over-the-shoulder hold, it leaves the title and all six faces untouched, and it looks like something a real publisher would wrap around a real book.

Prompt 8

GPT IMAGE 2.5 SUNBURST · BOOK EDIT · BELLY BAND Edit Image 1. Keep everything exactly the same: the four views of the black linen hardcover (front cover, angled view with spine, back cover, lying flat with red ribbon), the white background, the shadows, all existing typography, the arcads logos, the film strip with the six portraits and the numbers 01 to 06, the barcode. Add only one thing: a printed paper belly band (an obi band) wrapped horizontally around the lower part of the book, covering only the area below the film strip where "A FIELD MANUAL FOR PHOTOREAL AI FILMMAKING" sits, so the title, the arcads logo and all six portraits stay fully visible. The band is a strip of matte paper in the deep red of the ribbon bookmark, wrapping around the front, the spine and the back cover, with clean white condensed type centred on the front: "FREE TRIAL · 1,000 FREE CREDITS" and a smaller line beneath it: "FOR ALL NEW ARCADS CUSTOMERS" On the spine part of the band, only a small white arcads mark. The band shows a real paper edge, a slight gap where it lifts off the cover and soft shadow, following the perspective on the angled and lying-flat views; on the back cover view the band continues plain red with no text. Spell every word exactly as written. No other change anywhere.

OUTPUT

The final prop — VOL. 02 with the red offer band. This is @Image 8 in the film.
The final prop — VOL. 02 with the red offer band. This is @Image 8 in the film.

STEP 05 — SEEDANCE 2.5

Shooting the Film

Everything upstream exists to make this one generation easy. Seedance 2.5 runs in Reference to Video mode at 30 seconds, 4:3, 1080p, audio on, high bitrate, and we rendered two variants of the same prompt to choose from. The order of the nine references is the order of the @Image tags in the prompt, and each one is scoped: the lineup still controls location, blocking and grade but is explicitly not a first frame; the six charts restore faces; the book controls the cover; and a plain selfie rides along as face support for suspect four.

REFERENCE INPUTS — ATTACH IN THIS ORDER

@Image 1 — lineup still
@Image 1 — lineup still
@Image 2 — suspect 1
@Image 2 — suspect 1
@Image 3 — suspect 2
@Image 3 — suspect 2
@Image 4 — suspect 3
@Image 4 — suspect 3
@Image 5 — suspect 4
@Image 5 — suspect 4
@Image 6 — the witness
@Image 6 — the witness
@Image 7 — the detective
@Image 7 — the detective
@Image 8 — the book
@Image 8 — the book
@Image 9 — face support
@Image 9 — face support

SETTINGS

Seedance 2.5 · Reference to Video · 30 s · 4:3 · 1080p · audio on · high bitrate · two generations per run.

SEEDANCE 2.5 · VIDEO PROMPT · 30 S · 4:3 · 1080P

30-second cinematic multi-shot sequence, 4:3, 1080p, native dialogue and sound. Hard cuts between shots, real-time action, no slow motion.

A real police-station scene filmed on an iPhone, with the texture of an ARRI Alexa and ARRI Signature Prime lenses: low contrast, lifted blacks, soft highlight roll-off, creamy falloff, muted desaturated grade, fine grain. Handheld micro-shake, no smooth gimbal, except the lineup tracking shot. Lineup room: cold flat fluorescent light. Observation room: dark, lit only through the glass.

@Image 1: LOCATION, BLOCKING AND GRADE. The observation room, the one-way glass, the lineup wall, the four suspects holding cards 1 to 4, the old woman and the detective in the foreground. Not a first frame: start with gentle movement. @Image 2: SUSPECT 1, the platinum-ponytail woman. Character sheet: face, hair, outfit, silver hoops, nose stud. @Image 3: SUSPECT 2, the young man with black glasses. Character sheet. @Image 4: SUSPECT 3, the moustached man with cap and headphones. Character sheet. @Image 5: SUSPECT 4, the curly-haired man in the burgundy leather jacket. Character sheet: outfit and identity. @Image 6: THE OLD WOMAN. Character sheet. @Image 7: THE DETECTIVE. Character sheet. @Image 8: THE BOOK, with its deep red paper belly band. Use the front cover exactly as shown. @Image 9: SUSPECT 4 FACE SUPPORT. A real photo of the same man: use only his face to sharpen the likeness, bone structure, eyes, nose, moustache and goatee. Ignore clothing, room and selfie light. Character sheets: restore each face precisely in every shot; ignore grids, labels and grey backgrounds.

0.0 to 2.0s. Exactly the framing of @Image 1: behind the old woman and the detective, looking through the glass at the four suspects. The camera eases in slowly. 2.0 to 4.0s. Close-up of the detective from the glass side, the lineup reflected in his lenses. He points his notebook at the glass, dry and tired. {So, ma'am. Which ones are AI?} 4.0 to 5.5s. Close-up of the old woman leaning toward the glass, eyes narrowing, studying each face. 5.5 to 16.5s. Inside the lineup room, one continuous smooth lateral tracking shot on a dolly at chest height, gliding left to right without ever stopping, pushing in and drifting back out as it passes each suspect. 5.5 to 8.5s. Suspect 1 in close-up, chewing gum slowly, bored, eyes half-lidded, unbothered. {She will never figure it out.} 8.5 to 11.0s. The camera drifts in to an extreme close-up of Suspect 2's forehead: a single drop of sweat runs from his hairline down past his glasses. It eases back to the wide sliding angle; he lifts his chin, trying to stay cool, eyes flicking nervously. {It is pretty obvious I am real.} 11.0 to 14.0s. Suspect 3, jittery: sniffing, rubbing his nose, bouncing on his heels, scratching his neck, eyes darting, card trembling. He repeats fast under his breath. {Not me. Not me. Not me. Not me.} 14.0 to 16.5s. The camera reaches Suspect 4, relaxed, leaning slightly back. It pushes in very close to his mouth as one corner slowly pulls into a cool half smile. The track ends here. 16.5 to 18.0s. Hard cut: close-up of the old woman, quiet and certain. {One, two and three.} 18.0 to 20.5s. Lineup room, wide. A steel side door opens; the detective and two uniformed officers walk in. The detective stops at Suspect 4, tilting his head to the door. {Number four. You are free to go.} 20.5 to 23.0s. Officers cuff Suspects 1, 2 and 3: the woman rolls her eyes chewing, the young man stares in shock, the moustached man mouths silently. Suspect 4 drops his card and walks out. 23.0 to 26.0s. A police-station corridor at a property counter. A bored officer slides him a canvas tote bag. {Your belongings.} He takes it with a small nod. 26.0 to 30.0s. Over-the-shoulder from behind his right shoulder at head height as he walks down the corridor: he pulls the black hardcover book from @Image 8 out of the tote and holds it in front of his chest with both hands, cover up, looking down at it. The camera pushes in over his shoulder; he slows and stops. The cover reads exactly: ULTIMATE CINEMATIC REALISM BLUEPRINT, the arcads logo, six portraits numbered 01 to 06, and the red band: FREE TRIAL · 1,000 FREE CREDITS, FOR ALL NEW ARCADS CUSTOMERS, as on @Image 8. Final 1.5 seconds hold on the cover, sharp.

Diegetic sound effects only, no music. Fluorescent hum. Gum chewing on Suspect 1. Sniffing and shuffling feet on Suspect 3. <steel door buzzer and clunk> at 18.0s. <handcuffs clicking shut> at 21.0s, 21.6s and 22.2s. Footsteps on linoleum. <canvas tote sliding on the counter> at 23.5s. Soft rustle of the book from 26.5s.

male, mid-fifties, low dry American English voice, slightly gravelly, weary and unhurried.

female, seventies, soft American English voice, shaky but firm.

female, early twenties, flat bored American English voice with vocal fry, gum in her mouth, smug.

male, mid-twenties, American English, trying to sound calm, his voice cracking slightly.

male, thirties, American English, fast, breathy, high-strung, muttering.

male, forties, flat bored American English voice. SUSPECT 4 never speaks. Each line is spoken once, by its own character, in the shot where it appears.

Every face matches its character sheet in every shot; Suspect 4 also matches @Image 9. Card numbers never change. Suspect 1 chews gum throughout. Suspect 2 keeps his glasses on. Suspect 3 keeps cap, headphones and glasses on. Suspect 4 never wears sunglasses. The tracking shot moves left to right only and never stops. The book cover is identical to @Image 8.

music, extra spoken words, narration of the prompt, subtitles, captions, changed faces, swapped card numbers, sunglasses on Suspect 4, garbled book title or band text, stuttering camera, character sheet grids or labels, high contrast, crushed blacks, slow motion, watermark

STEP 06

Anatomy — Why It Reads as Real

  1. 01

    The contract

    “30-second cinematic multi-shot sequence, 4:3, 1080p … Hard cuts between shots, real-time action, no slow motion.” The first line tells Seedance it is cutting an edit, not drifting through one shot, and “real-time” removes the floaty slow motion that gives AI footage away.

  2. 02

    Two kinds of light

    “Lineup room: cold flat fluorescent light. Observation room: dark, lit only through the glass.” One sentence builds two rooms. The suspects live in the hard, honest light of a police room; the witness and detective live in the dark, lit only by what they are looking at.

  3. 03

    Scoped slots, plus a selfie

    “@Image 9: SUSPECT 4 FACE SUPPORT … use only his face to sharpen the likeness … Ignore clothing, room and selfie light.” Nine references only work if each one owns a single job. The charts carry outfits and identities; the selfie adds nothing but the face, so likeness improves without dragging a white tee into the lineup.

  4. 04

    One track that never stops

    “one continuous smooth lateral tracking shot on a dolly at chest height, gliding left to right without ever stopping, pushing in and drifting back out” Four suspects get four beats inside one move. The camera breathes in for each tell and out again, and the hard cut back to the witness lands twice as hard after eleven seconds without a cut.

  5. 05

    Character as behaviour

    “chewing gum slowly … a single drop of sweat runs from his hairline … sniffing, rubbing his nose, bouncing on his heels … one corner slowly pulls into a cool half smile” Nobody is described as guilty or innocent. Each suspect gets a physical tell, and the audience judges them exactly the way the witness does.

  6. 06

    Voices, and one silence

    “SUSPECT 4 never speaks. Each line is spoken once, by its own character, in the shot where it appears.” Six speakers get six distinct voice descriptions, and the one real person is the only one who never argues his case. The silence is the performance.

  7. 07

    Sound and insurance

    “<handcuffs clicking shut> at 21.0s, 21.6s and 22.2s … swapped card numbers … garbled book title or band text” Three cuffs, three timecodes, three arrests. The negative prompt blocks the failures that would break the joke: numbers that drift, a cover that will not read, a narrator reading the prompt aloud.

THE RESULT

Output — Real or Not

Seedance 2.5 · 30 s · 4:3 · 1080p with native dialogue. (Open this document in Google Docs to watch it play.)
Seedance 2.5 · 30 s · 4:3 · 1080p with native dialogue. (Open this document in Google Docs to watch it play.)

Audit it against the prompt: the opening frame matches the lineup still but already moves, the detective asks his question with the lineup in his lenses, the track glides past all four suspects without stopping, suspect one never stops chewing, the sweat drop runs, suspect three keeps muttering, my half smile arrives in close-up and I never say a word. The witness names one, two and three, three cuffs click on cue, and the final over-the-shoulder hold reads the title and the red band. If the cover ever wobbles, planar-track the GPT Image 2.5 Sunburst book render over the hold.

THE WITNESS ASKS THE QUESTION YOUR AUDIENCE IS ALREADY ASKING

The only real one never had to argue.

Swap the lineup for a jury box, a talent-show panel or a customs desk, the cards for name badges, the arrests for anything you like — the skeleton carries any spot-the-fake format. Cast from real references, lock every face into a chart, stage the room as a single still, design the prop that pays off the ending, then let one long reference-to-video generation direct the edit. Realism is not a filter you add at the end; it is a chain of small, scoped decisions, and every link in this blueprint is one of them.

Every model in this blueprint — GPT Image 2.5, GPT Image 2.5 Sunburst and Seedance 2.5 — lives in one studio. New Arcads customers start with a free trial and 1,000 free credits: enough to cast your own lineup and direct your first 30-second film.

arcads.ai →

WRITTEN, DIRECTED & PROMPTED BY DENIZ AKKABAK · @DENIZ.AKKABAK