REFERENCES @image1 = THE LOCATION: the Seoul street crossing at night with giant glowing LED billboards and neon signs, and the warm spotlight on the empty centre of the crossing. @image2 = THE DANCER. Face, wig, make-up and the complete stage costume (every piece, colour and accessory) come ONLY from @image2. @video1 = THE CHOREOGRAPHY ONLY. Copy the dance from @video1 exactly — every move, its timing, rhythm, arm positions, hair flips, steps and poses. Ignore the person, the clothes, the coloured background and the floor of @video1. CHARACTER LOCK The dancer from @image2 in every frame, wearing the complete costume and wig the whole time — the long ponytail and braid swing naturally with the moves and never change colour or length. One dancer only. Footage NOT mirrored. Real human proportions: the dancer's own height and build from @image2. SCENE A real K-pop music-video moment in the middle of the Seoul crossing from @image1: the dancer performs in the spotlight on the empty crossing; a few people far away on the sidewalks stop and film with their phones. Nobody walks in front of the dancer. All billboards and screens show ABSTRACT colourful graphics only — no readable text, brand names or logos. The asphalt is DRY and matte: no puddles, no water, no wet reflections. CAMERA A phone on a tripod at chest height, one continuous take with no cuts, almost static with a slight handheld drift, like @video1. The dancer is fully visible head to boots, in the centre, most of the time. PERFORMANCE Confident idol energy: sharp, clean, powerful moves exactly on the beat, a cool confident face with small smiles; the costume sequins sparkle in the spotlight. TEXT AND LOGOS No on-screen text, no captions, no logos anywhere. QUALITY Photorealistic, sharp face, correct anatomy — two arms, two legs, five fingers per hand — no merging, no extra limbs, feet never sliding or floating.

2 steps

MODEL: Seedance 2.5. Do not switch models. FORMAT: 9:16, 30 seconds, 1080p. Vertical frame. @image1 — THE SET: the empty moonlit garden with the oak trunk on the left and the empty rope swing, seen from this exact camera position. Keep this exact swing (one flat wooden plank on two thick ropes), the trunk, the hedge, the moon and the jack-o'-lanterns, and this exact side-on framing. @image1 shows no person; the only person in the film is @image2, sitting on this swing. @image2 — the person. Photocopy their face, facial structure, eye shape, skin tone, hairline, hairstyle, hair length, texture and colour, apparent age, build and body proportions exactly as shown in @image2. Do not beautify, slim, de-age, smooth, or restyle hair. Refer to them only as they/them. @image2 is identity only — not wardrobe. NATURAL LOOK: A real person filmed at night, not CGI, not a doll, not airbrushed. Real skin texture and pores. No porcelain skin, no beauty filter, no plastic face. FACE IN EVERY LOOK: @image2's own face is fully visible in every costume — no mask, no veil or web over the face, no face paint. No added makeup — no eyeliner, lipstick, contouring, glitter or false lashes — in any look. NATURAL MOVEMENT — CRITICAL: They sit on a real wooden rope swing with real weight. Hands hold the ropes. Knees and feet swing with the arc. The seat dips slightly at the bottom of each swing. Ropes taut, then slacken a little at the ends. Slow, human, slightly imperfect — not posed fashion holds, not frozen, not floating, not a mannequin. Hair and costume fabric follow the swing with real inertia. THE SWING — EXACT SETUP: One flat wooden plank hanging from two thick ropes tied to a high branch above the frame. The person sits ON the plank — hips and thighs on the seat, the seat directly under them, lower legs hanging below it. Never beside the seat, never in front of it, never floating next to it. The ropes hang on either side of their hips; both hands grip the ropes at chest height. Seen exactly side-on: the person faces the RIGHT edge of the frame in clean profile on every pass; one rope is in front of the body, one behind it. The swing moves only LEFT–RIGHT across the frame in a slow pendulum arc parallel to the camera. It never moves toward or away from the camera, never twists, and the person's size in frame stays the same. Only at 28.8–30.0s, when the swing has stopped, do they turn their head (not their body) to the camera. CAMERA: Fixed side view at swing height. Locked-off tripod. No camera movement, no cuts, no zoom. Vertical 9:16. The thick oak trunk is in the LEFT half of frame, near foreground, between camera and swing. The person lives ONLY on the RIGHT of the trunk for the entire film. They never appear on the left. They never change sides. ONE-SIDE TREE WIPE: The swing arcs slowly on the RIGHT toward the trunk, goes FULLY BEHIND the trunk (zero body/costume visible), then comes back out on the SAME RIGHT SIDE in ONE new Halloween costume. ONE hide behind the trunk = ONE new costume on the same side. Do NOT emerge on the left. Do NOT morph, dissolve, ghost, or double-expose clothes in the open. Do NOT change clothes while any part of them is visible. Do NOT make the swing fast. COSTUME SEQUENCE: 0.0–4.0s RIGHT SIDE: @image2 on the rope swing, fully visible right of the trunk, simple elegant dark outfit. Natural face from @image2. Slow human swing toward the trunk. 4.0–5.2s FULLY HIDDEN behind the trunk. The left of frame stays empty garden. No ghost. No person on the left. 5.2–9.5s RIGHT SIDE: emerge on the RIGHT as PUMPKIN ROYAL — burnt-orange costume with a pumpkin-shaped puffed body, autumn leaves and vines across the chest, small crown of tiny glowing jack-o’-lanterns. Slow natural swing on the right, then back toward the trunk. Hem flutters with inertia. 9.5–10.7s FULLY HIDDEN. Still nobody on the left. 10.7–15.0s RIGHT SIDE: emerge on the RIGHT as MOON MAGE — midnight-blue velvet cloak with silver stars, crescent-moon headpiece, lantern in one hand. Cloak follows the swing. 15.0–16.2s FULLY HIDDEN. 16.2–20.5s RIGHT SIDE: emerge on the RIGHT as BLACK CAT — sleek black cat suit, long tail, soft cat ears on top of the head, velvet gloves. @image2's own face uncovered — no cat mask, no face paint, no eye makeup. 20.5–21.7s FULLY HIDDEN. 21.7–25.5s RIGHT SIDE: emerge on the RIGHT as SPIDER MONARCH — black costume with silver spiderweb lace, a cobweb cape behind the head and shoulders (never over the face), jointed spider-leg shoulder pieces. Hair still exactly as in @image2 — do not restyle or bleach it. 25.5–26.7s FULLY HIDDEN. 26.7–28.8s RIGHT SIDE: emerge on the RIGHT as SCARECROW — patched burlap and plaid outfit, straw at sleeves and collar, worn straw hat, a crow on the swing rope. Natural @image2 face, no makeup. Arc eases like a real swing dying out. 28.8–30.0s RIGHT SIDE: swing slows to a stop on the right of the trunk. @image2 turns their head and looks into the camera with a slow knowing smile — natural @image2 face. The crow takes off. Hold. Hair stays exactly as in @image2 through every costume. Every costume is cut to fit @image2's own build and style — a long flowing robe, a fitted suit or anything in between, whichever suits them. ENVIRONMENT (unchanged): moonlit gothic garden at night, tall vertical composition. Huge full moon. Fog on the ground. Jack-o’-lanterns on the grass. Ivy stone wall. Bare branches. Falling autumn leaves. Teal moonlight, warm pumpkin candlelight. Photoreal live-action. AUDIO: creaking swing rope, soft wind, rustling leaves, a distant crow, a soft eerie music box. No dialogue. No voiceover. NO TEXT. NO SUBTITLES. NO END CARD. NO LOGOS. NO EXTRA PEOPLE. NO CAMERA MOVEMENT. NO CUTS. NO FAST SWING. NO LEFT-SIDE PERSON. NO VISIBLE MORPH. NO RUNWAY. NO STUDIO. KEY RULES: 9:16. Person only on the right of the tree, sitting ON the swing plank, seen side-on. Every clothes change happens fully behind the trunk, then they come back on that same side. Natural human swing motion.

REFERENCES @image1 = THE VISITOR. Face, hair, makeup and the complete costume come ONLY from @image1. Ignore the background of @image1. CHARACTER LOCK The visitor from @image1 in every frame where a person is visible, always the same face, makeup and costume. One person only. Footage NOT mirrored. FORMAT — REAL DOORBELL CAMERA FOOTAGE A home video-doorbell camera mounted next to a front door, looking down at the front porch from above: strong fisheye lens distortion. The WHOLE video is MONOCHROME infrared night vision — black-and-white / grey only, NO colour at all, flat cheap security-camera image with crushed blacks, the skin in its own tone from @image1 rendered in infrared grey, bright reflective eyes in the infrared, heavy digital noise, compression blocks, slightly low frame rate. It must look like real, cheap, low-quality doorbell footage — never cinematic, never colour graded. The camera NEVER moves — fixed position, one continuous take, no cuts. No on-screen text, no timestamp, no logos, no interface graphics. SETTING A suburban front porch at night: wooden steps leading up to the door, a railing, dead leaves on the steps, a dark yard and light fog behind, one weak porch light that flickers. The ground is dry. TIMELINE (12 s) 0.0–2.0 s — The porch is EMPTY. Nothing moves except the flickering porch light and drifting fog. 2.0–4.0 s — Suddenly, between two flickers of the light, the visitor from @image1 is STANDING perfectly still at the bottom of the steps at the edge of the frame — they did not walk in, they are simply there. Head tilted to one side at an unnatural angle, arms limp, staring up into the camera. 4.0–6.0 s — The footage stutters and drops frames three times; after each stutter the visitor is suddenly closer — halfway up the steps, then on the top step — without ever being seen walking. Still staring into the lens with an unsettling smile. 6.0–8.0 s — The visitor stands right in front of the door, close to the camera, face distorted by the fisheye. Very slowly they raise one finger toward the camera and clearly PRESS the small glowing doorbell button right next to the lens — the fingertip fills part of the frame as it pushes the button in. They never blink. 8.0–10.0 s — The porch light goes out. The image collapses into dark digital static with only the faint outline of the visitor's face and wide eyes visible in the noise. 10.0–12.0 s — The static clears in one frame: the visitor's face is PRESSED right against the lens, filling the whole frame, hugely distorted by the fisheye. The expression is TERRIFYING, not funny: lips CLOSED in a tight, unnaturally wide, motionless grin, completely still, dead glassy eyes wide open and unblinking, glowing in the infrared, staring straight into the lens. Then the head jerks sharply to one side in one unnatural snap with a sudden scream — and the image cuts to black. SOUND Real doorbell-camera audio: crickets, light wind and an electronic hum from the start; the porch light buzzing as it flickers. At 6.5 s a clear doorbell chime. At 8.5 s, in the static, a slow breathy whisper: "trick… or… treat…". At 10.3 s a sudden loud scream/hiss with a burst of digital distortion, then total silence on the black frame. No music, no other voices. QUALITY Photorealistic security-camera footage, believable and unpolished. No blood, no wounds, no gore. Correct anatomy, five fingers per hand.

CHARACTER LOCK — HIGHEST PRIORITY: The person in every shot is exactly the person in @image1. Preserve their facial identity, facial structure, eye shape and spacing, eyebrows, nose, mouth, exact skin tone, hairline, hairstyle, hair length, texture and colour, freckles and distinctive marks, apparent age, build and body proportions — identically, shot to shot. Preserve their gender presentation exactly as shown; do not masculinize, feminize or otherwise reinterpret it. Do not beautify, slim, de-age or restyle them. Do not add or remove bangs. Refer to them only as they/them in this prompt. ATTACHED IMAGES ARE MANDATORY AND FINAL: The attached images outrank every word written below them. Wherever an attached image and this text disagree, follow the image. ASSET MAP: @image1 — the person. Photocopy face, hair, skin, and the clothes on the still. Ignore the grey studio pad behind @image1 — that pad is not a set. Never freeze on the attached still as frame 0. Never show a checkerboard or transparency grid. @image2 — the product, exactly as photographed. Unchanged and not regenerated — same shape, colour, materials, lettering already on it, and hardware in every shot it appears. No extra unit, no invented logo. Ignore any empty white pad around @image2. SUBJECT LOCK: The person is @image1 in every shot, never replaced or restyled. HANDS: Natural hands, five fingers each, no deformation. Hands and fingertips stay in a slow, relaxed ASMR rhythm, never fast or sudden. SINGLE SUBJECT: One person only, no one else anywhere. PROP CONTINUITY: @image2 looks identical in every shot and every angle. Any lettering already on @image2 is part of the product. CAUSALITY LOCK: The soft pink-purple glow pulses only in response to each tap on @image2. Every sound has a visible cause: a tap, a slide, a rotation. PACING: The first second is already the hook, a crisp tap right at the microphone. Slow, intimate, smooth rhythm with clean cuts, no sudden movements. Never freeze on the attached still as frame 0. SCENE: A minimal dark room used as a streamer corner, vertical framing, mostly stable camera with a very slight handheld feel. The person sits at a desk with a microphone close to the camera. There is no visible main light source. The background is a clean, softly blended pink-purple gradient with ambient glow only, with subtle RGB edge light on the hair and shoulders. The person’s face is lit by a soft warm front light, creating a contrast between warm skin tones and the cool neon environment. @image2 is in the frame in every shot. English whispered dialogue, soft and breathy, one language only. 0-2s HOOK: Extreme close-up of fingertips tapping @image2 slowly twice right next to the microphone, crisp material-true taps, the pink-purple glow pulsing softly with each tap. The person’s face sits softly out of focus behind it, lit by the warm light. They whisper: “Listen to this.” 2-4s: The camera pulls back to a medium close-up. The person leans toward the lens, holds @image2 up into frame with both hands, and whispers: “Yeah. That sound.” 4-8s: Quick clean cuts: front, side and macro close-ups of @image2 as they slowly rotate it, tapping softly and lightly brushing along its surface, pink-purple reflections gliding along its edges. They whisper: “Look at these edges. So clean.” 8-11s: Extreme close-up of fingertips sliding along the surface of @image2, a brief pause, then one more slow tap. They whisper: “Hear that? That’s the real material.” 11-15s: @image2 is held close to the camera and tilted to catch the warm face light and the neon glow, the person gives a small satisfied nod and whispers: “That’s the one.” AUDIO: Highly detailed ASMR sound: fingertip tapping, light scratching, soft handling, with the sound matching the real material of @image2. The voice is a whisper. Very low ambient noise, an optional ultra-soft ambient tone, no music, no loud sounds. STYLE: Intimate, sensory, clean, minimal and visually soft, warm face light against a cool neon background. NEGATIVES: no subtitles, no on-screen text, no logos, no watermark, no other brands, no changes to @image1 or @image2, no flickering, no word “free”, no checkerboard, no studio-grey still as frame 0.

VERTICAL 9:16 from frame 0. 30 seconds. Photoreal cinematic comedy. Never open on a studio still or either attached photo as frame 0. No on-screen captions, logos, watermarks, or end cards except the two calendar shots described below. @image1 is the person. Photocopy their face, facial structure, eye shape, skin tone, hairline, hairstyle, hair length, texture and colour, apparent age, build and body proportions, makeup, and clothes exactly as shown in @image1. Do not beautify, slim, de-age, restyle hair, or wipe the makeup. Keep this same outfit in every shot. Do not add a different cape or costume. Refer to them only as they/them. @image2 is the product. Photocopy it exactly as shown: same shape, colour, materials, finish and any lettering. Unchanged. Free and unpackaged. Only one @image2 exists in every shot, always at real size. Do not restyle it. Do not put it in a box. Do not duplicate it on the bar. CONCEPT: they are so excited Halloween season has started that they rush out to celebrate with @image2 — and learn it is October 1. STYLE: photoreal gothic Halloween fairytale comedy. Candlelit old-world interiors, foggy cobblestone streets, moody blue night outside vs warm orange candlelight inside. Comedic timing. Smooth camera. CALENDAR RULES: Exactly two calendar shots, different. 1) OPENING (0.0-2.0s only): the revealed page is only the word OCTOBER. No day number. No date grid. 2) TAVERN wall calendar, including the LAST SCENE: a plain October month calendar with NO date marked, NO circle, NO highlight on any day. Completely unmarked. CAR / ARRIVAL: @image1 is the driver in every moving-car shot — in the seat, hands on the wheel. The car only moves while they are driving it. At the tavern: full STOP. Parked. Then @image1 opens the car door (door swings), steps out holding @image2, and walks through the tavern doors. Behind them the SAME car STAYS PARKED in that exact place. It does not creep, roll, reverse, or leave empty. 0.0-2.0s HOOK. Hand rips a calendar page: only OCTOBER, no date. Candles flare. Whip to @image1 delighted. Never start on the attached still. 2.0-5.5s GETTING READY. Bedroom. Their own clothes from @image1, mirror. @image2 on the dresser. Photocopy @image1 and @image2. No calendar. 5.5-7.5s GRAB. Picks up @image2, rushes out. 7.5-12.5s DRIVE. @image1 driving the vintage black convertible, visible in the seat, @image2 buckled on the passenger seat. Mansion, lightning, cobblestone. 12.5-18.5s ARRIVAL. Car with @image1 at the wheel slows and STOPS in front of the timber-framed tavern. Full stop, parked. @image1 opens the car door, steps out with @image2. The parked car stays in that same place. They walk through the tavern doors. 18.5-21.5s TAVERN. Empty bar, no Halloween decorations. Places ONE @image2 on the counter. Photocopy @image2. No calendar yet. 21.5-24.5s REVEAL. Elderly bartender, distinct from @image1. Points at a paper calendar on the tavern wall. Close-up: an October calendar with NO date marked or circled. 24.5-28.0s PUNCHLINE. Smile melts. Bartender, English: "Halloween is in thirty days." @image1 sits by the single @image2: "...I'll wait." Two mugs. 28.0-30.0s FINAL. Wide empty tavern: @image1 and bartender at the bar, ONE @image2 between them. The wall calendar is visible and has NO date marked. Slow push-in on @image2. Hold. AUDIO: playful spooky orchestral until the tavern, silence at the reveal. SFX: paper rip, candles, coat, car engine that STOPS, car door, footsteps, tavern doors, cricket. Dialogue English, clear. NO SUBTITLES. NO TEXT except the two calendars. NO WATERMARK. NO LOGOS other than lettering already on @image2. NO EXTRA PEOPLE besides @image1 and the bartender. NO DUPLICATE @image2. NO EMPTY CAR DRIVING. NO MARKED DATE ON THE WALL CALENDAR.

REFERENCES @image1 = THE LOCATION: the empty gothic ballroom decorated for a winter ball — frosted white branches, hanging crystal icicles, gothic windows, cold blue light and haze. The whole video happens here. @image2 = THE DANCER. Face, hair, make-up and the complete outfit come ONLY from @image2 — keep every piece of the outfit exactly as in @image2. @video1 = THE CHOREOGRAPHY ONLY. Copy the dance of the main dancer in @video1 exactly — every move, its timing, rhythm, head snaps, arm positions, hand shapes and steps — performed by ONE dancer. Ignore everything else in @video1: other people, blurred areas, the room, the floor and the side borders. CHARACTER LOCK Exactly ONE dancer: the person from @image2, in every frame, in the same outfit the whole time. No other people anywhere in the hall. Footage NOT mirrored. STYLE Deadpan gothic attitude: a calm, serious, expressionless face, steady stare, never smiling — the charm comes from the sharp, slightly strange moves done with total seriousness. High energy: the dancer is already dancing in frame 0 and never stands still. Every move is BIG, sharp and full-arm, hitting hard and exactly on the beat — snappy head jerks, wide arm swings, fast hand flicks; hair and fabric swing and fly with every turn. CAMERA One continuous handheld phone shot at chest height, no cuts, slight natural drift, full 9:16 frame with no borders. MEDIUM SHOT: the dancer is in the centre and framed from the knees up the whole time, the body filling most of the frame height — never a small figure in a wide shot. At 14.0 s the dancer leans toward the lens and the face fills the upper half of the frame. GROUND Dark MATTE wooden dance floor, completely DRY — no puddles, no water, no shine, no reflections. ENDING (14.0–15.0 s) The dancer leans toward the camera and freezes with a dead-serious stare straight into the lens. TEXT AND LOGOS No on-screen text, no captions, no logos anywhere. QUALITY Photorealistic, sharp face, correct anatomy — two arms, two legs, five fingers per hand — no extra limbs, feet never sliding or floating.

2 steps

REFERENCES @image1 = THE DANCER. Face, hair, makeup, mask and the complete Halloween costume (every piece, colour and accessory) come ONLY from @image1. @image2 = THE LOCATION: the city place where the dance happens, with its light, buildings and mood. Ignore any people standing in the centre of @image2. @video1 = THE CHOREOGRAPHY ONLY. Copy the dance from @video1 exactly — every move, its timing, rhythm, body positions, jumps, low lunges and the final walk toward the camera. Ignore the person, the clothes, the room and the floor of @video1. CHARACTER LOCK The dancer from @image1 in every frame, wearing the complete costume the whole time — any mask, hat, cape or accessory stays on and in place. One dancer only. Footage NOT mirrored. SCENE A real influencer moment in the middle of the city location from @image2: the dancer performs in an open spot while real life goes on — people walk past behind the dancer in both directions, a few stop and film the dancer with their phones, some smile and point. Nobody walks in front of the dancer or touches them; passers-by stay in the background, slightly out of focus. Any screens, signs or billboards show ABSTRACT colourful graphics only — no readable brand names, text or logos. The pavement is DRY and matte: no puddles, no water, no reflections on the ground. CAMERA A phone on a tripod at chest height, one continuous take with no cuts, almost static with a slight handheld drift; the dancer is fully visible head to feet most of the time. Around 13–15 s the camera slowly pushes in a little and then settles back. At the end the dancer walks toward the camera until they are framed from the waist up. CHOREOGRAPHY (follow @video1, 23 s) 0.0–0.8 s — standing still, hands at the head with two fingers up like little ears. 0.8–2.0 s — wide lunge to the left, one arm bent up high, then a jump with the arms crossed above the head. 2.0–4.5 s — lunge to the right with the arms in a V, then the arms frame the head, wrists crossed above the head, body waves. 4.5–6.0 s — deep, wide, low lunges left and right, arms swinging. 6.0–8.0 s — low crouch with sharp arm flicks and torso contractions. 8.0–8.6 s — big back arch with the arms over the head, then drop down to one knee. 8.7–10.0 s — rise, arms spread wide like wings, small jumps — any cape or loose fabric of the costume opens and flies. 10.0–12.0 s — turn to the side, sharp hand gestures in front of the face, stepping. 12.0–13.2 s — deep forward bend, back to the camera, torso swinging. 13.2–15.0 s — rise with one arm reaching high and a twist. 15.0–15.6 s — explosive jump with one arm thrown up, wide landing, loose parts of the costume flying. 15.7–17.6 s — arms crossed over the face, then flung up, then folding down. 17.7–23.0 s — straightens up and walks slowly toward the camera with the hands on the waist, head tilts, hair flips, confident influencer attitude, ending close to the lens with a smirk. COSTUME PHYSICS Any cape, skirt, sleeves or loose fabric of the costume swing, flare and trail naturally with every turn, jump and lunge; the hair swings. The costume never covers the hands or blocks the moves. TEXT AND LOGOS No on-screen text, no captions, no logos anywhere. QUALITY Photorealistic street video, sharp dancer, correct anatomy — two arms, two legs, five fingers per hand — no extra limbs, no merging with passers-by, feet planted with real weight.

REFERENCES @image1 = THE DARK HOUSE: the empty abandoned family home at night, broken window, cold blue moonlight, dust, family photos scattered on the floor. The location only — the people come from @image2, @image3 and @image4. @image2 = PERSON A as a ZOMBIE — the zombie look for PART 1: grey-green pale skin, dark veins, cloudy milky-white eyes, tangled dusty hair, torn dirty t-shirt. Still clearly the face of @image3. No blood, no wounds, no gore. @image3 = PERSON A as they really are (normal, happy, alive). Face, features and hair of PERSON A come ONLY from @image3. @image4 = PERSON B. Face, features and hair of PERSON B come ONLY from @image4. CHARACTER LOCK Exactly two people in the whole video: PERSON A and PERSON B. Nobody else. Footage NOT mirrored. Correct anatomy, two hands with five fingers each. Cinematic film look, shallow depth of field, real emotion. PART 1 — THE DARK HOUSE (0.0–9.2 s). Cold blue moonlight, slow and heavy, a movie scene cut into short shots. SHOT 1 (0.0–1.5 s) — Extreme close-up of PERSON B's face: eyes full of tears, breathing hard, the top of a raised pistol soft and blurred at the bottom of the frame, held in front of the chest, shaking. SHOT 2 (1.5–2.5 s) — Wide shot of the dark house like @image1: PERSON A as the zombie stands still at the far end of the room, CLEARLY LIT by a beam of cold moonlight from the broken window — the pale zombie face, tangled hair and torn t-shirt clearly visible, not a dark silhouette. SHOT 3 (2.5–4.0 s) — Close-up of the zombie's face (@image2): milky-white eyes, head tilting slowly, a low raspy breath. SHOT 4 (4.0–5.5 s) — Extreme close-up of PERSON B's face: one tear rolls slowly down the cheek, lips trembling; PERSON B whispers, voice breaking: "I'll always love you." SHOT 5 (5.5–6.5 s) — Medium shot from the front: PERSON B slowly lowers the pistol and lets it fall from the hands; it drops and hits the floor (clearly visible). SHOT 6 (6.5–7.5 s) — Same medium shot, hands now EMPTY: PERSON B slowly spreads both arms straight out to the SIDES at shoulder height, fully extended like a cross, palms open and facing forward, fingers relaxed and open — surrendering, "come to me". The arms stay wide open like this through SHOTS 8 and 9 until the zombie leaps into them. Never point the hands forward, never make fists or thumbs-up. SHOT 7 (7.5–8.2 s) — Close-up: the zombie snarls with the mouth open and lunges straight toward the lens. SHOT 8 (8.2–8.6 s) — Close-up of PERSON B, arms still wide open: closes the eyes, accepting. SHOT 9 (8.6–9.2 s) — Wide side shot: the zombie sprints across the room and LEAPS into PERSON B's open arms. The cut happens at the exact moment their bodies meet, mid-leap. 9.2 s — MATCH CUT ON THE LEAP: the same leap continues without a break, but the world changes from the dark house to bright sunshine. PART 2 — THE HAPPY MEMORY (9.2–18.0 s). A perfect holiday on a Maldives island: white powder sand, crystal-clear turquoise water, palm trees leaning over the water, overwater villas in the distance, warm golden sun. PERSON A is completely normal again — exactly the face, skin and hair of @image3, no zombie make-up at all. Both wear light, fully covered summer clothes. SHOT 10 (9.2–11.5 s) — Side shot at the waterline: PERSON A lands the leap in PERSON B's arms, legs wrapped around, arms around the neck; PERSON B catches and spins PERSON A around once, both laughing, foreheads touching. SHOT 11 (11.5–14.0 s) — They run knee-deep into the turquoise water and playfully splash water at each other with their hands, laughing, droplets sparkling in the sun. SHOT 12 (14.0–18.0 s) — Golden sunset: they sit close together on the sand, seen from behind and slightly to the side, PERSON A's head on PERSON B's shoulder, arms around each other, watching the sun sink into the sea. The camera slowly drifts back and holds this calm moment until the end. SOUND A sad, emotional piano and soft strings play under the whole video, swelling at the leap (9.2 s) and turning warm in PART 2. PART 1: quiet house ambience, PERSON B's breathing and the whispered line, the zombie's raspy breath and snarl, the thud of the pistol on the floor. PART 2: waves, laughter and splashing under the music. The ONLY spoken words in the whole video are the whisper "I'll always love you." in SHOT 4. TEXT AND LOGOS No on-screen text, no captions, no logos anywhere. QUALITY Photorealistic, cinematic, emotional, sharp faces that always match the references.

2 steps10

One continuous handheld first-person shot, no cuts, like real found-footage of someone exploring a haunted house with a phone. The person filming is NEVER visible — no body, no shadow, no hands, no phone in the frame. Nobody else is in the house. The video starts on the given image. Calm, calm, calm… then the most violent, terrifying scare. 0–6 s: CALM. The camera slowly walks through the dark abandoned room toward the painting at the far end, past the sheet-covered furniture, floorboards creaking, dust floating in the moonlight. Quiet breathing of the person filming. 6–10 s: CALM. The camera arrives in front of the painting and STOPS, framing the painted figure's face and upper body. The person filming quietly admires it, a soft impressed whisper: "wow…". The painting is completely still, a beautiful old oil portrait; the candles flicker. Long, uncomfortable silence. 10.0 s: SUDDENLY the painted figure's eyes snap straight into the lens and the face twists into a horrifying expression — the jaw drops unnaturally low, the mouth stretches far too wide, the eyes go pitch black, the face warps and distorts like melting oil paint. 10.5–13 s: In one explosive instant the figure LUNGES OUT of the painting at full speed, the canvas tearing, wet oil paint stretching off them in thick strings, both clawed hands shooting at the lens to GRAB the camera. A loud, piercing, inhuman CACKLE fills the room. The distorted face rushes right into the lens; the camera is grabbed, shakes violently and spins. 13–15 s: The camera falls, the candles go out, total darkness — only the inhuman cackle echoing and fading. Sound: total near-silence at first — creaking floorboards, soft breathing, a faint whispered "wow…"; then at 10 s a sudden loud horror sting, the tearing of canvas, a piercing inhuman cackle, the violent grab and a thud, then echoing cackle in the dark. No music. Photorealistic found-footage horror, genuinely terrifying, sudden and fast. No blood, no gore, no text, no logos.

3 steps

REFERENCES @image1 = THE PERSON. Face, facial features, head shape, eye colour, skin tone, natural hair and own clothes come ONLY from @image1. @image2 = THE LOOK. Use @image2 ONLY for the makeup (colours, shapes, marks, lip colour), the hair styling / wig / headpiece and the costume. Never take the face, facial features, age or body of anyone shown in @image2 — the face is always the face of @image1. @image3 = THE LOCATION. From the first transition (2.6 s) to the end, the video happens in the place shown in @image3, with its light and mood. Ignore any people in @image3. The attached audio = the music track under the whole video. CHARACTER LOCK The same person from @image1 in every shot, clearly recognisable through every makeup stage. Footage NOT mirrored: every one-sided makeup mark stays on the same side of the face as in @image2. LOCATION AND CAMERA 0.0–2.6 s: the person's own cozy bedroom. From 2.6 s to the end: the location from @image3 in every shot. INFLUENCER SELFIE STYLE: the person films themself with the phone's front camera; framing is never wider than waist-up — the full body is never shown. Vertical 9:16, handheld phone camera with energetic shake, sudden push-ins into tight close-ups of the face and fast pull-backs, quick whip-tilts on the strongest beats, everything moves on the beat of the music. One continuous take with no cuts: the look and the location change only while something covers the lens (described below), and every transition is ONE smooth, unbroken motion — the camera never stops moving through it. Never a static frame. PERFORMANCE Influencer music-video energy, dramatic and confident, like a creator who owns the camera. The person SINGS the lyrics below straight into the lens in a breathy, eerie, rhythmic half-sung voice, on the beat, with clear lip-sync, and never stops moving, even bigger and more dramatic than a normal performance: a new move on every beat — sharp head snaps, hair flips, shoulder shimmies, hands framing the face, a finger pointed at the lens, leaning in close then snapping back, slow eerie body rolls with hard hits on the strong beats. Big dramatic expressions: arched eyebrows, wide eyes, smirks, a slow wicked smile. Theatrical, playful, haunting. TIMELINE (15 s) 0.0–3.0 s — STAGE 1, NATURAL, THE HOOK: @image1 exactly as in the photo, no makeup change, own clothes, hair loose, sitting at a vanity table in their own cozy bedroom — warm lamp light, a ring light glow, makeup brushes and palettes on the table; the phone camera is where the vanity mirror would be, so no mirror is ever visible. Frame 0: the person lunges into an extreme close-up, eyebrows raised, points one finger straight at the lens and starts singing line 1 immediately; then snaps back with a hair flip and a shoulder shimmy on the beat. At 2.6 s the person shoves one open palm flat onto the lens — the palm fills the frame — and PUSHES the camera away: the camera flies backward from the extreme close-up into a wider shot, and the person is now in the location from @image3. 3.0–7.0 s — STAGE 2, BASE: as the palm leaves the lens the face now has only the base layer of the @image2 makeup, clearly visible — the full-face skin base colour applied strongly and evenly over the whole face, and the eye shading — no other details, natural lips, own clothes, hair loose. The camera rushes back in. The person sings line 2 while patting the cheeks with a large fluffy powder brush, with a head snap and a hip sway between strokes, twirling the brush on the beat. At 6.6 s the person grabs the camera with both hands, pulls it into an extreme close-up of one eye, then SHOVES it away — the camera flies backward into a wider shot. 7.0–11.0 s — STAGE 3, FULL MAKEUP: as the camera flies back, the face wears the COMPLETE makeup from @image2 — every colour, mark and line — and the person is ALREADY in the middle of smearing the lipstick on: at the very first frame of this stage the lipstick bullet (a real lipstick tube, not a brush) is already pressed to the lips, gliding across them in one confident stroke. Still own clothes, hair loose, no costume yet. The camera rushes back in close. The person finishes the stroke with a smirk, tosses the lipstick over the shoulder, then sings line 3 with hands framing the face and a slow wicked smile, leaning into the lens. At 10.6 s the person throws the sheer veil from @image2 (or, if the costume has no veil, a long sheer piece of the costume fabric) toward the camera and the veil sweeps across the whole lens. 11.0–15.0 s — STAGE 4, FULL LOOK: when the veil settles, in a chest-up selfie shot, the person now wears the ENTIRE look from @image2: the same makeup, the same hair styling / wig / headpiece, the same costume (seen from the chest up). The camera pushes back in to a close-up selfie framing. Dramatic dance on the beat: a slow body roll, sharp head snaps, one arm sweeping up, shoulders hitting every beat. The person sings line 4, laughing the three "ha" syllables with one sharp shoulder jerk each. 13.9–15.0 s — FINAL POSES (mandatory): the person stops dancing and strikes three sharp, dramatic model poses on the last three beats — chin down with eyes up into the lens; one hand framing the face; a slow turn of the head over the shoulder — all in close-up selfie framing (face and shoulders), and FREEZES in the last pose with a slow knowing smile straight into the lens while the lights flicker out. LYRICS (sung — ONLY these lines, in this order, each line sung once) Line 1 (0.4–2.5 s): "Powder pale, I'm fading fast," Line 2 (3.8–6.2 s): "Paint the cracks, forget the past," Line 3 (8.2–10.4 s): "Lace and veil, the clock strikes one…" Line 4 (11.6–13.8 s): "Ha-ha-ha — the haunting's begun!" SOUND The soundtrack IS the attached audio at FULL VOLUME, loud and clearly audible from 0.0 to 15.0 s — the music is LOUDER than the voice, the beat and bass are the main sound — with the person's singing voice softer on top of it. No other voices, no other music, no sound effects. TEXT AND LOGOS No on-screen text, no captions, no logos anywhere. QUALITY Photorealistic, cinematic, sharp face in every stage, natural skin texture under the makeup, correct anatomy: two hands, five fingers each, the brush and lipstick held naturally. No blood, no wounds, no mirrors, no other people.