SEEDANCE 2.5 | TEXT TO VIDEO | SINGLE CONTINUOUS TAKE | 22s | 1080x1920 | 9:16 | 30fps | 1080p SCENE AND LOCATION A hidden desert oasis slot canyon deep in a red sandstone plateau on a bright midday. A narrow, deep gorge with steep, very high walls of warm red-orange sandstone, wind-carved into smooth ripples and streaked with dark desert varnish, with hardy green shrubs and small palms clinging to ledges. At the bottom is a crystal-clear natural oasis pool. The shallow edges glow bright turquoise and aqua, fading to deep sapphire blue in the center. The water is so transparent that the bottom is fully visible: pale sandy floor, smooth boulders, and two long sun-bleached driftwood logs lying diagonally across the bottom. One spring-fed waterfall pours from a crack high on the canyon wall into the pool, making white foam where it hits the water. Tall date palms and desert scrub line the canyon rim. Above is a vivid clear blue desert sky with a few small white clouds. Hard, bright midday sunlight with strong contrast, heat-shimmer in the far distance, and sun sparkles on the water. THE JUMPER A person seen only from their own first-person view. Tanned bare feet and athletic shins, and loose terracotta-orange swim shorts with a simple cream geometric pattern and white drawstrings. No shirt, no shoes, no gloves. Their hands and forearms appear naturally in frame when they move. CAMERA AND REALISM First-person POV from a GoPro strapped to the forehead. It looks like real action-cam footage, not a cinematic camera. 16mm-equivalent ultra-wide lens with slight barrel distortion, and saturated but natural color. Use real action-cam behavior throughout: small head bob and micro-jitter from breathing and body weight, rolling-shutter wobble on fast moves, quick auto-exposure shifts between bright sky and the shaded canyon, a small sun flare on upward tilts, light sensor noise in the shadows, and wind buffeting that shakes the frame during the fall. The cliff is about 35 meters high. Gravity, body motion and water physics are fully realistic. 0-2s | THE EDGE (real time) POV looking straight down from the top of the sandstone cliff. Bare feet stand on the rough, warm rock ledge at the bottom of frame, with the shorts visible at the lower edge. Both hands hover forward at the sides of frame. The turquoise pool, the driftwood logs and the waterfall are far, far below. Toes curl over the edge and small grit crumbles off. A hand tosses an orange-and-white striped inflatable tube out over the drop; it tumbles down and lands as a tiny speck on the water far below. Wind, distant waterfall roar, upbeat chill electronic music starts. 2-4s | THE COUNTDOWN (real time) A quick nervous tilt down at the drop, a sharp exhale, then two fast steps back from the edge, with the ledge bouncing in frame. A hyped voice: “Three, two, one!” 4-7s | THE TAKEOFF (natural slow motion, about 50% speed) Two running steps forward, then the last foot pushes off the ledge in slow motion. Rock grit kicks up and floats in the sunlight, the cliff edge slides away underfoot, arms swing forward and out wide, and the whole turquoise oasis opens up below. The slow motion looks like a real high-frame-rate action-cam clip, not a frozen hang. Deep bass swell, stretched wind. 7-13s | THE FALL (real time, fast and continuous) The speed ramps smoothly back to normal and the stomach drops. 7-9s: The frame dips and wobbles, arms flail slightly for balance at the edges of frame, feet kick at the bottom of frame. A loud yell: “Woooooo!” 9-11s: The sandstone wall, its ripples and streaks, and clinging desert shrubs streak upward past the lens, and the waterfall rushes by with spray catching the sun. Wind roars louder and the frame shakes from air buffeting. 11-13s: The glowing turquoise water rushes up and fills the frame, and the orange tube and driftwood logs grow fast. Arms snap tight to the body and legs lock straight for a feet-first pencil dive. The music builds to the drop. 13-14s | IMPACT A hard feet-first hit into the water and a white explosion of bubbles swallows the lens. Loud plunge splash, deep impact whoosh, and the music goes heavily muffled underwater. 14-17s | UNDERWATER (short, energetic, real time) A violent swirl of silver bubbles clears to reveal crystal-clear blue water, bright sun shafts cutting down from the surface, the sun-bleached driftwood and pale sandy bottom below, and small fish darting away. Hands push into frame for one strong breaststroke pull, then fast kicks straight up toward the bright glowing surface, with bubbles streaming past the lens. No slow sinking. 17-19s | SURFACE BREAK The camera bursts through the surface, water droplets sliding across the lens, and bright sunlight floods in. The music snaps back to full clarity on the beat drop. A hand grabs the orange-and-white tube floating right beside it. 19-22s | THE REWARD POV bobbing on the turquoise water holding the tube, then a sharp tilt up revealing the towering sandstone cliff, the tiny ledge jumped from far above, the palm trees on the rim and the vivid blue sky, with the waterfall splashing to one side. An out-of-breath, triumphant voice: “Yeah, that was insane!” Faint distant cheering. Music plays clean to the final frame. AUDIO Continuous upbeat chill electronic track: a bass swell on takeoff, a fast build through the fall, muffled underwater, then a clean drop on surfacing. Real ambient layers: wind that grows with speed, fabric flapping, heavy breathing, waterfall roar, splash, bubbles. Dialogue clear over the music. KEEP One continuous take, first-person only. The same shorts, feet and hands throughout, and the same canyon layout with one waterfall, turquoise pool and submerged logs. The slow motion lasts only 3 seconds at takeoff. The fall is fast, real time and builds speed without pausing. Every beat is quick and punchy. AVOID Extra slow motion beyond the takeoff, hovering or floating mid-air, smooth cinematic gimbal movement, cuts to third person, showing the jumper’s face, extra people, distorted hands or extra fingers, morphing cliffs, changing water color, text, subtitles, watermarks.

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. No face changes, no identity drift. No other people share their face. ATTACHED IMAGES ARE MANDATORY AND FINAL: If an avatar image is attached, that is the person — reproduce them exactly, in every single frame, and never invent a different face, hairline, hair colour, skin tone, body or age. ASSET MAP: @image1 — the character. @image1 is face, hair, skin and identity only. Do not copy a black V-neck long-sleeve top, a gold snake pendant, layered gold chains, a gold bangle, a crossed-arms studio pose, or a white seamless backdrop from the still. WARDROBE (locked for the whole clip): a fitted black tank top and heather-gray sweatpants. Bare arms. No gold jewelry. SET: a real lived-in bathroom, waist-up. Soft daylight from a window or vanity. A mirror off-camera, a simple sink edge, a toothbrush cup, a small hand towel. Stable objects. Natural smartphone exposure, realistic skin texture, no cinematic grading, no beauty filter. CAMERA: real smartphone beauty UGC. Waist-up, eye-level, very subtle natural handheld movement only. Face sharp and stable. One continuous take. No cuts. PRODUCT: one mascara wand and matching tube, generic black, consistent shape and size. No brand marks, no readable label. HANDS: only the visible hand holding the mascara wand. Always anatomically correct with five fingers, naturally proportioned. Do not create another hand, arm or additional limb. Do not duplicate the wand. START: already in the bathroom from frame 0, already holding the mascara wand near their upper lashes, already speaking. Never a frozen studio pose. Never the attached still as an opening plate. Never cut to the attached still. Tell the following story in exactly 10 seconds, same person throughout. 0.0–2.0s HOOK. Waist-up. Wand already at the upper lashes. They speak immediately, casual and unscripted, exact lip sync: "Okay wait, I actually really love this one." Tiny handheld sway. 2.0–5.5s APPLY. They lightly brush the wand along the upper lashes, slow and careful. Exact lip sync: "It lengthens my lashes so easily, it doesn't clump at all." Natural blink. Face stays @image1. 5.5–8.0s CHECK. They lower the wand a little, glance at the result in a small hand mirror or just off-camera, then look back to camera. One deliberate slow blink. Exact lip sync: "And it makes my eyes look way more awake" 8.0–10.0s APPROVE. Small pleased expression, slight nod. Wand still in the same hand, tube stable. Exact lip sync: "without looking overdone." Hold. Designed end, not a fade. PERFORMANCE: casual beauty-creator recommendation. Slightly persuasive but still natural. Subtle, confident, pleased. Small natural head motion. Normal irregular blinking. Minimal gestures. AUDIO (native, synced speech, no BGM, no voiceover, no extra narrator): they speak the lines above in a natural speaking voice, plus quiet bathroom room tone. Accurate mouth movement. REALISM: natural pores and skin texture, realistic hair strands, natural lash movement, no clumped or spider lashes, accurate eyelid and blink mechanics, natural lip sync, stable bathroom objects, no object duplication, no changing anatomy. NEGATIVE: duplicate hand, extra arm, extra fingers, floating hand, malformed hand, duplicated product, identity drift, altered face, changing hairstyle, changing clothing, gold snake necklace, black V-neck long sleeve, crossed-arms studio pose, white seamless backdrop, warped anatomy, mirror artifacts, beauty filter, plastic skin, AI glow, robotic movement, exaggerated gestures, flickering, CGI, clumped lashes, spider lashes, distorted eyes, smudged eye makeup, on-screen text, subtitles, watermark, logos, extra people.

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. No face changes, no identity drift. No other people share their face. ATTACHED IMAGES ARE MANDATORY AND FINAL: If an avatar image is attached, that is the person — reproduce them exactly, in every single frame, and never invent a different face, hairline, hair colour, skin tone, body or age. The attached images outrank every word written below them: wherever an attached image and this text disagree, follow the image. ASSET MAP: @image1 — the character (the florist). @image1 is face, hair, skin and identity only. Wardrobe is florist-shop work clothes from the scene (simple work shirt or apron) — do not copy an olive-green quarter-zip, a white crew t-shirt, or the lounge interior from the still. SINGLE SUBJECT — HIGHEST PRIORITY: Only this one person appears throughout. No other people or hands ever enter the frame. CAUSALITY LOCK: The open shop door is already visible and slightly ajar from shot 1, so the gust of wind that swings it open further and scatters the stems has a clear, visible source. Nothing moves without a shown cause. PROP CONTINUITY: One partly-finished bouquet, one pair of shears, and a small pile of loose stems and petals on the counter all exist from shot 1. Nothing new appears later. HANDS: hands are always anatomically correct with five fingers each, naturally proportioned, in every shot they appear. Tell the following story in exactly 3 shots, same character throughout, where shot 2 is a genuinely different event from shot 1 — not a bigger version of the same action. Clear visual beginning, twist and ending — never ambiguous or faded-out. STORY: The character is already mid-work at the counter, the shop door slightly ajar behind them, trimming stems for a bouquet with fast, confident, practiced snips of the shears — no measuring, clearly skilled, placing each stem into the arrangement in one fluid motion. A sudden gust swings the door open wider, and it visibly scatters the loose stems and petals on the counter, sweeping several off onto the floor and knocking the bouquet slightly off-balance in its vase. The character’s hands move quickly and precisely — completely different motion than the cutting before — gathering the scattered stems and petals off the counter and floor in fast, sure sweeps, tucking them back into the arrangement and straightening the vase, finishing with the bouquet looking fuller and better than before. They push the door shut with one hand, step back, and smile at the finished arrangement. STYLE: Warm, grounded flower-shop realism, soft natural light from the doorway, shallow depth of field, camera close on hands and face throughout, a sudden gust of motion for the scatter, settling into a steady frame for the recovery. Vertical 9:16. Continuous real motion, no freeze-frame, no cutaways. No on-screen text, no subtitles, no watermark, no title card. AUDIO (native, synced, no BGM, no voiceover, no narration): the snip of shears, quiet shop ambience, a sudden gust and the door swinging with a creak, the scatter of stems and petals hitting the counter and floor, quick decisive gathering sounds, the door clicking shut, and a satisfied exhale. SHOT 1 (0-6s): Open already inside the work — no static establishing beat. Character trimming stems with fast, confident snips and placing them into the bouquet, door slightly ajar behind them. SHOT 2 (6-12s): The turn — a different event, not a repeat. A gust swings the door open wider, scattering stems and petals off the counter and onto the floor, knocking the bouquet off-balance. SHOT 3 (12-18s): The resolution. Quick, precise gathering — a different hand motion than the cutting — restores the scattered stems to the arrangement, the vase is straightened, the door is pushed shut. They step back and smile at the finished bouquet. A clear, decisive ending, not a fade-out.

FLOWER MORPH — The Bush Becomes a Person @image1 walks up and touches a big flowering bush — and the blossoms spiral upward and knit into @image2, who steps out in a flowing gown. A magical two-avatar transformation, cinematic 15s. Plug in ANY two avatars.

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. No face changes, no identity drift. No other people share their face. ATTACHED IMAGES ARE MANDATORY AND FINAL: If an avatar image is attached, that is the person — reproduce them exactly, in every single frame, and never invent a different face, hairline, hair colour, skin tone, body or age. The attached images outrank every word written below them: wherever an attached image and this text disagree, follow the image. ASSET MAP: @image1 — the character (the frightened protagonist). Photocopy @image1 for face, hair, skin, identity, and the wardrobe on the still (burgundy long-sleeve V-neck satin dress, layered gold chain with a gold snake pendant, thin gold bangle). Do not restyle hair. Do not invent a different dress or jewellery. Ignore any beige pad around the still — that pad is not a set. If any other person appears (the tactical soldier), they are never named, aged or gendered in the text — described only by their role and what they do. Only @image1's face is locked as an identity. Extra people in this story are beats, not copies of @image1. HANDS: hands are always anatomically correct with five fingers each, naturally proportioned, in every shot they appear. Create a cinematic photorealistic action-thriller sequence set in a modern downtown city surrounded by tall skyscrapers, at dusk with the city lights just coming on. SHOT 1 (0-3s): Extreme close-up of the character’s frightened face — the person exactly as in @image1 — wet eyes, realistic skin texture, dramatic dusk lighting, intense emotional expression. SHOT 2 (3-7s): Cut to a high aerial drone shot flying between dense skyscrapers, looking down at a busy city street with cars, pedestrians, traffic markings and realistic urban detail. The camera moves rapidly forward between the buildings with cinematic motion blur. SHOT 3 (7-12s): Cut to a dynamic POV shot from inside a moving dark sedan, focusing on the side-view mirror. In the mirror reflection, a heavily armored futuristic black tactical vehicle drives aggressively alongside the car, with a tactical soldier visible inside — described only by role and action, never named or identified. Keep the mirror reflection realistic and sharply detailed while the surrounding city moves with natural motion blur. SHOT 4 (12-16s): Fast-paced aerial and street-level shots showing the armored vehicle chasing the sedan through the downtown streets, including a brief stretch through an underpass tunnel lit by sodium lights. Realistic smoke, dust, damaged vehicles and cinematic atmosphere. SHOT 5 (16-20s): End with a dramatic top-down aerial shot looking directly at a city intersection as the sedan moves through traffic and a powerful explosion suddenly erupts nearby, producing a bright orange fireball, thick gray smoke, sparks and debris. STYLE: Ultra-realistic physics, realistic vehicle movement, cinematic camera transitions, natural lighting, shallow depth of field, high detail, realistic motion blur, blockbuster Hollywood action-movie cinematography, photorealistic, HDR, anamorphic lens, dramatic color grading, intense suspenseful atmosphere. AUDIO (native, synced, no BGM, no voiceover, no narration): engine roar and tire screech, wind rushing past the drone shots, the tactical vehicle’s engine snarling alongside, distant sirens, the explosion’s deep boom with debris scattering, and the character’s frightened breathing in the close-up. No on-screen text, no subtitles, no watermark, no logos. Vertical 9:16. Exactly 20 seconds.

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. No face changes, no identity drift. No other people share their face. ATTACHED IMAGES ARE MANDATORY AND FINAL: If an avatar image is attached, that is the person — reproduce them exactly, in every single frame, and never invent a different face, hairline, hair colour, skin tone, body or age. The attached images outrank every word written below them: wherever an attached image and this text disagree, follow the image. ASSET MAP: @image1 — the character. @image1 is face, hair, skin and identity only. Wardrobe is a fitted olive-green utility jacket, brown trekking trousers, sturdy hiking boots, and a small tan canvas backpack. Hair and jacket move naturally with the mountain wind throughout. Do not copy a black V-neck top, a gold snake pendant, layered gold chains, a gold bangle, a crossed-arms studio pose, or a white seamless backdrop from the still. SINGLE SUBJECT — HIGHEST PRIORITY: Only this one person appears throughout. No other people, faces or hands ever enter the frame. HANDS: hands are always anatomically correct with five fingers each, naturally proportioned, in every shot they appear. ENVIRONMENT: a dramatic high-altitude mountain landscape — misty granite ridgelines, alpine meadows dotted with wildflowers, a still turquoise alpine lake, distant snow-capped peaks, and thin cloud drifting across the rock faces. Crisp, elevated, adventurous atmosphere. LIGHTING AND CINEMATOGRAPHY: warm golden-hour sunlight cutting through drifting mist, soft highlights on rock and water, natural shadows, realistic skin texture, cinematic depth of field, subtle lens flare, smooth camera movement, natural motion blur, high-end travel-film color grading. Tell the following story in exactly 15 seconds, same character throughout, already in motion from frame 0 — never a frozen studio pose. Clear visual beginning, middle and ending. SHOT 1 (0.0-3.0s): RIDGELINE. Open already in motion — no frozen pose. Aerial wide of a misty mountain ridge, fog already drifting across the rock, wind already moving the character’s jacket and hair as they stand at the edge of an overlook, back toward the camera, looking out over the peaks. The camera slowly descends and moves closer. SHOT 2 (3.0-6.0s): RIDGE TRAIL. Smooth tracking shot as they walk along a narrow rocky trail through an alpine meadow, boots finding footing on stone, wildflowers brushing past their legs. Jacket and hair move naturally in the wind. Camera follows from behind and slightly to the side. SHOT 3 (6.0-9.0s): CLOSE-UP. Intimate close-up as they turn toward the camera with a soft, natural smile, windblown hair across the face. Warm sunlight on the skin, misty peaks blurred behind. Photocopy @image1 face. SHOT 4 (9.0-12.0s): LAKESIDE. Side-profile beside the still alpine lake, looking toward the distant snow-capped peaks, eyes gently closing to feel the mountain wind. Camera slowly circles as sunlight creates a soft golden rim light on the hair. SHOT 5 (12.0-15.0s): ENDING. Wide shot walking toward a scenic rocky viewpoint over the mountain range as the sun sets in soft golden, peach and pink tones. Camera pulls backward and upward, revealing the full ridgeline and valley below. End on a peaceful, dreamy frame — a clear last beat, not a fade-out. MOOD AND STYLE: dreamy, peaceful, adventurous, elegant, premium cinematic travel-film atmosphere. Ultra-realistic, photorealistic, cinematic 1080p, realistic human anatomy, consistent facial identity, natural facial expressions, realistic mist and cloud movement, realistic wind, natural lighting. No artificial CGI appearance, no distorted anatomy, no identity changes, no unnatural movements. AUDIO (native, synced, no BGM, no voiceover, no narration): wind gusting across the ridge, distant faint bird calls, boots on rock and gravel, fabric moving in the wind. No on-screen text, no subtitles, no watermark, no logos. Never cut to the attached still.

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. No face changes, no identity drift. No other people share their face. ATTACHED IMAGES ARE MANDATORY AND FINAL: If an avatar image is attached, that is the person — reproduce them exactly, in every single frame, and never invent a different face, hairline, hair colour, skin tone, body or age. The attached images outrank every word written below them: wherever an attached image and this text disagree, follow the image. ASSET MAP: @image1 — the character (the driver). Photocopy @image1 for face, hair, skin, identity, and the wardrobe on the still (burgundy long-sleeve V-neck satin dress, layered gold chain with a gold snake pendant, thin gold bangle). Do not restyle hair. Do not invent a different dress or jewellery. Ignore any beige pad around the still — that pad is not a set. If any other person appears (the tactical soldier), they are never named, aged or gendered in the text — described only by their role and what they do. Only @image1's face is locked as an identity. Extra people in this story are beats, not copies of @image1. HANDS: hands are always anatomically correct with five fingers each, naturally proportioned, in every shot they appear. Create a cinematic photorealistic action-thriller sequence set in a modern downtown city at night, surrounded by tall skyscrapers, told as one clear cause-and-effect chain — not disconnected spectacle shots. SHOT 1 (0-4s): Close-up on the character driving, exactly as in @image1, one hand on the wheel, the other resting protectively on a small sealed metal case on the passenger seat. They glance anxiously into the rearview mirror, already uneasy — establishing that they are carrying something valuable and already suspect they’re being followed. Realistic skin texture, dramatic night lighting, tense expression. SHOT 2 (4-7s): Cut to a high aerial drone shot flying between the skyscrapers, following the character’s car through the busy night street below, then pulling back to reveal a heavily armored black tactical vehicle several cars behind, closing the distance with clear intent. The camera moves rapidly forward with cinematic motion blur, establishing that this vehicle is specifically hunting the character’s car. SHOT 3 (7-11s): Cut to a POV shot from inside the character’s car, focused on the rearview mirror. In the reflection, the tactical vehicle pulls up alongside, a tactical soldier visible inside, gesturing for them to stop. The character grips the case tighter and accelerates instead, confirming they won’t give it up. Keep the mirror reflection sharply detailed while the city blurs past. SHOT 4 (11-15s): Fast-paced aerial and street-level shots of the chase through downtown — the character weaving through traffic to shake the tactical vehicle, which rams a parked delivery truck while trying to cut them off, spilling cargo and forcing its own path toward a roadside fuel tanker idling at a loading zone. Realistic smoke, dust, and cinematic atmosphere. SHOT 5 (15-20s): The tactical vehicle, still trying to close the gap, clips the fuel tanker at speed. A powerful explosion erupts behind the character’s car — a bright orange fireball, thick gray smoke, sparks and debris — visibly caused by that collision, not appearing at random. The character’s car speeds clear of the blast radius; a final shot through the rearview mirror shows the fireball and stalled tactical vehicle receding into the distance as the character exhales, adrenaline still high but visibly safe. STYLE: Ultra-realistic physics, realistic vehicle movement, cinematic camera transitions, natural night lighting, shallow depth of field, high detail, realistic motion blur, blockbuster Hollywood action-movie cinematography, photorealistic, HDR, anamorphic lens, dramatic color grading, intense suspenseful atmosphere. AUDIO (native, synced, no BGM, no voiceover, no narration): engine roar and tire screech, wind rushing past the drone shots, the tactical vehicle’s engine snarling alongside, the crash into the truck, the tanker’s explosion with a deep boom and scattering debris, and the character’s tense breathing settling into relief in the final shot. No on-screen text, no subtitles, no watermark, no logos. Vertical 9:16. Exactly 20 seconds.

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. No face changes, no identity drift. No other people share their face. ATTACHED IMAGES ARE MANDATORY AND FINAL: If an avatar image is attached, that is the person — reproduce them exactly, in every single frame, and never invent a different face, hairline, hair colour, skin tone, body or age. The attached images outrank every word written below them: wherever an attached image and this text disagree, follow the image. ASSET MAP: @image1 — the character. Photocopy @image1 for face, hair, skin, identity, and the wardrobe exactly as worn on the still. Add a simple day bag they carry. Do not restyle hair. Do not invent different clothing. Ignore any beige pad around the still — that pad is not a set. If any other person appears (the security guard, the HR associate, the department head), they are never named, aged or gendered in the text — described only by their role and what they do. Only @image1's face is locked as an identity. Extra people in this story are beats, not copies of @image1. HANDS: hands are always anatomically correct with five fingers each, naturally proportioned, in every shot they appear. SETTING & CAMERA: One continuous modern office: entrance → security desk → badge gate → glass corridor → open office → whiteboard aisle → the character’s own desk. Morning light, hard floors, realistic office furniture. A white PTZ security camera is already mounted above/right of their desk and must remain there throughout. 0-25.5s: continuous third-person tracking, usually 1.5m behind the character’s shoulder. Realistic follow-lag, whip-pans to threats, no hidden cuts or teleportation. 25.5s: hard cut to a fixed frontal medium close-up showing the character, their desk, and the PTZ camera. HUD: a simple colour-based detection indicator top-center (green → yellow → red, no numeric readout). Minimap bottom-left; STAMINA/COMPOSURE top-left; SPRINT/CROUCH/INTERACT/SLIDE bottom-right icons only, no numbers. Opening text: “LATE — 12 MIN” / “OBJECTIVE: REACH YOUR DESK UNDETECTED.” 0-7s — SECURITY GUARD: The character sprints through the entrance with their bag. The guard notices them. They hide behind a large planter instead of confronting the guard, dropping low as the guard approaches with a radio. The indicator flares toward red. When the guard looks away, they quietly slip out, badge through the gate, and continue. The guard notices the closing gate but does not chase. The indicator eases back toward green. 7-13s — HR CART: HR pushes a tall cart loaded with folders toward the character. They accidentally brush the glass, then quickly crouch behind the cart and move with it, staying hidden as HR checks both sides. They circle behind the cart when HR looks away, then break cover and run once the corridor is clear. HR never clearly sees them. The indicator peaks red, then drops back toward green. 13-18s — DEPARTMENT HEAD DISCOVERY: The department head appears at a T-junction and notices the character’s reflection through the glass. They turn toward the character. A solid red “!” flashes exactly twice above the department head. Only afterward, the indicator turns full red: “ALERT.” The department head calls, “Hey! Hold on!” The character turns and says, “Oh, crap!” The department head follows their route. 18-23.5s — WHITEBOARD ESCAPE: The character races through the open-office aisle, grabs a wheeled whiteboard, and pulls it diagonally across the department head’s path. The department head brakes before hitting it and immediately circles around. The character squeezes through the remaining gap between the whiteboard and a desk using a low side-slide, keeping their bag tight to their body, then immediately gets up and runs. The department head stays behind them. 23.5-25.5s — FALSE SUCCESS: The character reaches their desk, drops into the chair, puts the bag underneath, and starts typing while catching their breath. The department head searches the rear aisle but cannot see them. The indicator falls to green. Green “MISSION COMPLETE” appears. The character smiles with relief. 25.5-30s — CCTV TWIST: At 26.2s, the pre-installed PTZ camera physically whirs, pans and tilts toward the character, then its red indicator illuminates. They slowly look up toward it and freeze. Remove “MISSION COMPLETE.” Show “CCTV DETECTED,” followed by red “MISSION FAILED.” They mutter, “You’ve got to be kidding me.” Hold their defeated expression, the red-lit camera, and the failure HUD until the end. PHYSICS & RESTRICTIONS: real footsteps, shoe friction, breathing, bag movement, glass reflections, radio, badge gate, cart/whiteboard/chair wheels, keyboard sounds, and PTZ motor. No teleportation, clipping, frictionless movement, clones, identity changes, distorted hands, plastic skin, impossible reflections, or furniture hitting people. Only the department head pursues the character. The guard and HR associate never chase. The CCTV camera must physically turn; the HUD must not replace the physical camera action. AUDIO (native, synced, no BGM, no voiceover, no narration besides the spoken lines): hard-floor footsteps, bag fabric, radio chatter, badge-gate beep, cart wheels, glass tap, the department head’s “Hey! Hold on!”, the character’s “Oh, crap!”, whiteboard wheels, keyboard keys, PTZ motor whir, then “You’ve got to be kidding me.” No subtitles, no BGM, no extra logos. Vertical 9:16. Exactly 30 seconds.

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. No face changes, no identity drift. No other people share their face. ATTACHED IMAGES ARE MANDATORY AND FINAL: If an avatar image is attached, that is the person — reproduce them exactly, in every single frame, and never invent a different face, hairline, hair colour, skin tone, body or age. The attached images outrank every word written below them: wherever an attached image and this text disagree, follow the image. ASSET MAP: @image1 — the character (the florist). @image1 is face, hair, skin and identity only. Wardrobe is florist-shop work clothes from the scene (simple work shirt or apron) — do not copy an olive-green quarter-zip, a white crew t-shirt, or the lounge interior from the still. SINGLE SUBJECT — HIGHEST PRIORITY: Only this one person appears throughout. No other people or hands ever enter the frame. CAUSALITY LOCK: The open shop door is already visible and slightly ajar from shot 1, so the gust of wind that swings it open further and scatters the stems has a clear, visible source. Nothing moves without a shown cause. PROP CONTINUITY: One partly-finished bouquet, one pair of shears, and one held stem all exist from shot 1. The loose stems and petals that scatter in shot 2 come only from the bouquet already established. Nothing new appears later. PACING: All human motion in this scene is unhurried and deliberate — no rushed, sped-up or frantic movement at any point, including while picking scattered items back up. Allow brief, natural pauses between actions. The gust of wind itself is a quick natural event, but the character’s own movements stay calm and controlled throughout, even during the recovery. HANDS: hands are always anatomically correct with five fingers each, naturally proportioned, in every shot they appear. Tell the following story in exactly 3 shots, same character throughout, where shot 2 is a genuinely different event from shot 1 — not a bigger version of the same action. Clear visual beginning, twist and ending — never ambiguous or faded-out. STORY: The character is already at the counter, the shop door slightly ajar behind them, holding a single flower stem firmly in one hand at a marked length. They position the open shears around the stem, pause a beat, then close the blades slowly and deliberately, cutting cleanly through it in one visible, unhurried snip — the cut stem clearly falling free into their palm. They place it into the bouquet with calm, practiced care. A sudden gust swings the door open wider, and it visibly scatters several loose stems and petals off the counter onto the floor, knocking the bouquet slightly off-balance in its vase — a clear, different event from the calm cutting before. The character sets down the shears, kneels, and picks the scattered stems and petals back up one at a time, unhurried, straightening each one before returning it to the arrangement. They right the vase, step back, and take a slow, satisfied look at the finished bouquet. STYLE: Warm, grounded flower-shop realism, soft natural light from the doorway, shallow depth of field, camera close and steady on hands and face throughout, holding on the single deliberate cut in shot 1, a brief quick beat for the gust itself, then settling into an unhurried, steady pace for the calm recovery. Vertical 9:16. Continuous real motion, no freeze-frame, no cutaways, no time-lapse or sped-up motion anywhere. No on-screen text, no subtitles, no watermark, no title card. AUDIO (native, synced, no BGM, no voiceover, no narration): quiet shop ambience, the single clean snip of the shears through the stem, a pause, a sudden gust and the door swinging with a creak, stems and petals scattering onto the counter and floor, then calm, unhurried gathering sounds one stem at a time, the door clicking shut, and a slow, satisfied exhale. SHOT 1 (0-8s): Open already inside the work — no static establishing beat. Character holding a single stem firmly in one hand, positioning the shears around it, and cutting through it in one clear, deliberate, unhurried snip, then placing it into the bouquet. The stem must be visibly present and visibly cut, not an empty gesture. SHOT 2 (8-16s): The turn — a different event, not a repeat. A gust swings the door open wider, scattering stems and petals off the counter and onto the floor, knocking the bouquet off-balance. This is the one quick natural moment in the scene. SHOT 3 (16-24s): The resolution. Calm, unhurried gathering — the character kneels and picks up each scattered stem one at a time, straightens it, and returns it to the arrangement, then rights the vase. They step back and take a slow, satisfied look at the finished bouquet. A clear, decisive ending, not a fade-out.