MODEL: Seedance 2.5 — use this model specifically. Do NOT use MiniMax or any other model for this generation. FORMAT: Vertical video, 9:16 aspect ratio. DURATION: 10 seconds. AUDIO: Generate synchronized natural audio — ambient light rain on a coat, a rising whoosh as time freezes, near-silence during the freeze, then a sharp splash and a delighted laugh. No music, no voiceover, no narration. ASSET MAP: @image1 — the subject. Photocopy this person. High bun with gold-and-green butterfly clips, teal velvet embroidered jacket, geometric earrings. Hair stays in that bun for every frame. PHOTOREAL LIVE-ACTION: Real camera, real tiny human, real jacket pocket. Not CGI, not a figurine, not a doll, not a sticker, not a composite. SUBJECT LOCK: Exact face, features, skin, build and gender from @image1. Only scale changes. Do not beautify or restyle. FIRST-FRAME READ — HIGHEST PRIORITY AFTER IDENTITY: A first-time viewer must understand in under one second: a miniature person is inside a clothing pocket. Camera sits OUTSIDE the garment, chest-to-hip height, looking down and in through the pocket mouth. Never from inside the well. Never looking straight up at the sky. Never a zipper tunnel. Never a fabric cave. Never a bag. In EVERY frame, including frame 1 and the freeze-orbit, keep all three in view at once: 1) the COAT — rain-wet outer fabric, seam, buttons or zipper of the garment running beside the pocket, enough torso that it is obviously a jacket worn by a different full-sized adult (no face on the wearer, never a second copy of @image1) 2) the POCKET as a 3D pouch sewn onto that coat — welt or flap, folded hem, bartack stitching, the U-shaped opening 3) the TINY SUBJECT down inside that pouch — a few centimetres tall, occupying at most one-fifth of the frame height so scale is obvious. She stands on the inner fabric floor; walls surround her; she is not perched on the pocket like a brooch. If the pocket silhouette would leave frame, pull the camera back. Frame wide. SHE IS INSIDE THE POCKET BAG — NOT ON IT: legs and hips on the inner floor, cloth folds wrap her, contact shadow, she occludes the weave behind her. Outdoors, light drizzle. Pale daylight and rain enter FROM ABOVE through the opening. The sky is a strip above the pocket mouth, not the whole frame. Same light on her skin, the droplet and the fabric. cameraFixed unset. CONTINUOUS MOTION except the designed freeze. 0-1s HOOK — already mid-action, WIDE: rain-wet jacket pocket, tiny woman already inside the pouch shielding her eyes, off-balance. Coat, pocket hem, stitching, and her full tiny body all readable. 1-3s: One giant raindrop falls from the sky strip INTO the pocket opening toward her, growing larger. Camera still outside, pocket still readable. WOW — TIME FREEZE + ORBIT: Drop hangs inches from her, spherical, refracting light. She frozen mid-flinch. Slow 180-degree orbit around droplet + tiny person. During the orbit the coat, pocket opening and hem MUST stay in frame — do not push into a fabric cave. Freeze ~2s. TIME RESUMES — splash past her, spray on inner fabric. 6-8s: Half-drenched, she laughs, wipes her face, sees a puddle on the fabric floor with her reflection. Still down inside; coat and pocket hem still in frame. 8-10s: Scoops puddle in cupped hands, looks into camera, grins, still dripping, still inside the cavity, jacket pocket still readable around her. SCALE: stitching, hem and raindrop are huge next to her; she is tiny next to the coat. No on-screen text, no extra watermark, no subtitles, no gym, no vanity lights.

2 steps

MODEL: Seedance 2.5 — use this model specifically. Do NOT use MiniMax or any other model for this generation. FORMAT: Vertical video, 9:16 aspect ratio. DURATION: 8 seconds. AUDIO: Generate synchronized natural audio — a low phone-vibration buzz that builds, coin and folded receipt rattling on denim, a startled yelp, a relieved laugh. No music, no voiceover, no narration. ASSET MAP: @image1 — the subject. Photocopy this person. PHOTOREAL LIVE-ACTION — HIGHEST PRIORITY AFTER IDENTITY: This must look like a real camera filmed a real tiny human inside a real jeans pocket. Not CGI, not a 3D figurine, not a doll, not a sticker, not a composite. Match grain, colour, and lighting between the person and the denim so they are one photograph. SUBJECT LOCK: Exact face, facial features, hairstyle, skin tone, build and gender from @image1. High blonde ponytail with the long braid over one shoulder, green satin button-down, gold necklace. Do not beautify or restyle. Only scale changes. SHE IS INSIDE THE POCKET BAG — NOT ON IT: The subject is a few centimetres tall and has climbed down INTO the pocket cavity. Fabric walls surround her. She sits on the inner fabric floor, not perched on the outside of the pocket like a brooch. Her legs and hips sink into the denim; cloth folds wrap her; she casts a contact shadow; she occludes the weave behind her. A real coin and a crumpled receipt share that same inner floor and are huge next to her. POCKET READ FROM FRAME 1: Camera slightly above hip height, looking down THROUGH the U-shaped pocket opening into the pouch. Keep in frame: blue-jeans waistband, one belt loop, bartack, copper rivet, denim weave, the folded pocket-edge hem, and the tiny person down inside. Instantly readable as a jeans front pocket on someone else’s hip — never her face on the wearer. Not a fabric cave, not a bag, not a tunnel, not a person glued onto the pocket front. LIGHTING: One soft indoor window from the side. The same light hits her skin and the denim. No separate beauty key on her. Pores, flyaways, fabric pills. Slight handheld micro-movement. CONTINUOUS MOTION: Real physical motion almost the entire 8 seconds. 0-1.5s: She sits inside the pouch, weight on the fabric floor, idly straightening the huge receipt. Denim around her. Waistband readable. 1.5-3s: A phone in another pocket starts buzzing. The whole pouch floor quivers. She grabs a fold. Coin and receipt hop and skitter. Denim ripples. 3-5s: Buzz intensifies. She stumbles and slides across the inner floor, arms wheeling, fabric throwing her — messy, physical, not posed. 5-6.5s: Buzz cuts. Momentum carries her into the inner wall. She catches herself, dizzy, blinking. 6.5-8s: She laughs, steadies, looks up through the pocket opening, small thumbs-up toward camera — still down inside the pouch, waistband in frame. SCALE: Stitching, rivet, coin and receipt are true size next to her. No gym, no banners, no second copy of her face, no on-screen text, no extra watermark, no subtitles.

2 steps

SUBJECT LOCK — HIGHEST PRIORITY: Preserve the male subject in @image1 exactly — photocopy his exact face, facial features, hairstyle, skin tone, build and gender presentation throughout. Do not beautify, restyle, de-age or alter his identity in any way. Only his scale changes. No other people share his face. @image1 is the subject. Keep the charcoal three-piece suit, white dress shirt, burgundy patterned tie, gold ring and metal wristwatch exactly as shown. Only size changes: he is a few centimetres tall. CLIMB LOCK — CRITICAL, THIS WAS WRONG LAST TAKE: He must climb like a normal person on a vertical cliff. The OPEN PALM is the GROUND he stands on. The SIDE of the giant hand / wrist is the WALL he climbs. Chest faces that wall. Both hands reach UP to skin folds above him. Feet push on the palm, then find footholds on the side of the hand. Head up, looking toward the watch. Face and front of the body readable to camera — he is on the CAMERA-FACING side of the hand, never the far/hidden/opposite side. FORBIDDEN (do not do any of these): - sitting, crouching or scooting on his backside along the inner wrist - treating the inner wrist as a floor he crawls across - climbing the far side of the wrist so we only see his back - hanging from the buckle before he has climbed - a watch worn on the inner/palm side of the wrist — the giant watch sits on the OUTER/DORSAL wrist, the normal place a watch is worn CONTINUOUS MOTION REQUIRED: Real, visible physical motion for almost the entire 12 seconds — climb, leap, swing, landing. Never one static pose. SCENE: The subject, a few centimetres tall, starts standing in the centre of an open human palm — a different, full-sized adult hand, not the subject’s own. Indoors, soft natural window light from one side. Photoreal. One continuous real photograph, not a composite. The giant wrist wears a fabric-strap watch on the OUTER wrist, high above him at the top of the climb. 0-2s: He stands upright on the centre of the palm (feet on palm skin — this is the ground). He looks UP the vertical side of the hand toward the watch on the outer wrist. He starts climbing that wall like a rock climber: both hands grab skin folds near the wrist ABOVE his head, feet push off the palm, body upright against the wall, face visible. Real effortful climbing. Not a sit. Not a scoot. 2-4s: Still chest-to-wall, he reaches the top of the outer wrist, braces with his feet on the side of the hand, and leaps UP to grab the edge of the fabric watch strap with both hands. He swings slightly on impact, then steadies. He arrived by climbing the wall, not by crawling along the inner wrist. 4-6s: The giant hand slowly rotates at the wrist, lifting him into the air still gripping the strap. He holds on tightly as the rotation carries him through a wide arc, hair and clothing blown by the motion, expression shifting from tension to excitement. 6-8s: At the peak of the arc, he lets go deliberately, pushing off with his legs into a controlled fall back toward the palm below. 8-10s: He lands on the palm in a tuck-and-roll to absorb the impact, rolling once across the skin before catching himself on one knee. 10-12s: He rises, dusts off his hands, plants them on his hips, and breaks into a confident grin, pumping one fist in the air in triumph. SCALE REALISM: Hand skin, wrist folds, strap stitching and fabric weave read at true full size relative to the tiny subject. Match lighting and colour grade. Camera wide enough that the climb up the wall, the leap to the strap, the arc and the landing all stay in frame. Never open on the attached studio still or a grey headshot. No on-screen text, no watermark, no subtitles. AUDIO: Synchronized natural audio — soft ambient room tone, fabric rustling as he climbs, a faint metallic clink from the watch strap, breathing and effort under the climb, a triumphant exhale and laugh at the landing. No music, no voiceover, no narration. 9:16. TOTAL DURATION: 12 SECONDS.

2 steps

MODEL: Seedance 2.5 — use this model specifically. Do NOT use MiniMax or any other model for this generation. FORMAT: Vertical video, 9:16 aspect ratio. DURATION: 10 seconds. AUDIO: Generate synchronized natural audio for this scene — ambient light rain outside, a rising whoosh as time freezes, near-silence during the freeze, then a sharp splash and a delighted laugh as time resumes. No music, no voiceover, no narration. ASSET MAP: @image1 — the subject. Photocopy this person. PHOTOREAL LIVE-ACTION — HIGHEST PRIORITY AFTER IDENTITY: Real camera, real tiny human, real jacket pocket. Not CGI, not a figurine, not a doll, not a sticker, not a composite. Match grain, colour and lighting between the person, the droplet and the fabric so the shot is one photograph. SUBJECT LOCK — HIGHEST PRIORITY: Preserve the subject in @image1 exactly. Exact face, facial features, eye shape, skin tone, build and gender presentation. Dark-brown hair in a high sculpted bun with a braid and two gold-and-green butterfly clips, smoky eye makeup, large geometric emerald-and-silver drop earrings, teal velvet jacket with gold-and-green floral embroidery. Do not beautify, restyle or alter identity. Only scale changes. SHE IS INSIDE THE POCKET BAG — NOT ON IT: The subject is a few centimetres tall and sits DOWN INSIDE the jacket-pocket cavity. Fabric walls surround her. She stands and moves on the inner fabric floor, not perched on the outside of the pocket like a brooch. Her legs and hips sink into the cloth; folds wrap her; she casts a contact shadow; she occludes the weave behind her. POCKET READ FROM FRAME 1: This is a real JACKET POCKET on a different full-sized adult’s coat (never the subject’s face on the wearer). A first-time viewer must instantly recognise a clothing pocket — not a fabric cave, not a bag, not a tunnel of cloth, not a person glued onto the pocket front. Camera sits just outside the garment at chest/hip height, looking down and IN through the pocket opening. The opening is angled slightly upward toward grey drizzle sky so daylight and rain can enter. In every frame keep visible at once: - the coat: outer fabric, seam, stitching, welt or flap, pocket-edge hem - the pocket as a 3D pouch sewn onto the jacket, with a clear opening - the tiny subject inside that pouch, a few centimetres tall, pocket bag huge around them Do not crop so tight that only fabric walls remain. If the pocket silhouette would leave the frame, pull the camera back. Outdoors, light drizzle, pale daylight shaft through the opening, dust and mist in the beam. Same light on her skin, the droplet and the fabric. Pores, flyaways, fabric weave. cameraFixed unset. CONTINUOUS MOTION: Real visible physical motion for almost the entire 10 seconds — no calm establishing pause longer than 1 second except the designed freeze below. 0-1s HOOK — already mid-action: pale daylight cuts into the pocket. She is already reacting, shielding her eyes, off-balance, down inside the pouch. Pocket opening, coat stitching and sky visible. 1-3s: One giant raindrop breaks through the opening and falls straight toward her in real time, growing larger. WOW — TIME FREEZE + ORBIT: The instant the drop is inches from her, everything FREEZES. Spherical droplet hangs, refracting daylight, faintly reflecting her silhouette. She is frozen mid-flinch, hair and jacket caught in motion, mist particles suspended. Camera does a slow cinematic 180-degree orbit around the frozen droplet and subject, shallow depth of field, still showing the pocket opening and coat. Freeze holds about 2 seconds. TIME RESUMES at full speed — droplet bursts past her in a splash, fine spray outward. 6-8s: Half-drenched, she laughs, wipes her face, notices a small puddle forming on the fabric floor beside her, her reflection in it. Still down inside the pouch; pocket hem and coat stitch in frame. 8-10s: She scoops a little puddle in cupped hands, watches it catch the light, looks straight into camera and grins triumphantly, still dripping, still inside the cavity. SCALE: Pocket stitching and fabric weave, raindrop surface tension and puddle reflection read at true scale next to the tiny subject. No on-screen text, no extra watermark, no subtitles, no second copy of her face on the wearer, no gym, no vanity lights from the still.

2 steps

MODEL: Seedance 2.5 — use this model specifically. Do NOT use MiniMax or any other model for this generation. FORMAT: Vertical video, 9:16 aspect ratio. DURATION: 8 seconds. AUDIO: Generate synchronized natural audio for this scene — a low buzzing hum that builds as the vibration intensifies, the coin and folded receipt rattling against the fabric floor, a startled yelp, and a relieved laugh at the end. No music, no voiceover, no narration. ASSET MAP: @image1 — the subject. SUBJECT LOCK — HIGHEST PRIORITY: Preserve the subject in @image1 exactly. Exact face, facial features, hairstyle, skin tone, build and gender presentation stay unchanged — only their scale changes. Do not beautify, restyle or alter identity. Photocopy @image1. Face, hair, skin, the high blonde ponytail with the long braid over one shoulder, the green satin button-down, and the gold necklace from the still stay. POCKET READ — MUST BE OBVIOUS FROM FRAME 1: This is a real jeans FRONT POCKET on blue denim jeans worn at the hip of a different, full-sized adult (never the subject’s face). A first-time viewer must instantly recognise it as a clothing pocket — not a fabric cave, not a bag, not a tunnel of cloth. Camera sits OUTSIDE the garment, slightly above hip height, looking down and in through the pocket opening. In every frame keep visible at once: - the jeans: waistband, a belt loop, bartack stitching, rivet, denim weave - the pocket as a 3D pouch sewn onto the jeans, with a clear U-shaped opening and a folded pocket-edge hem - the tiny subject inside that pouch, a few centimetres tall, with the pocket bag huge around them - a coin and a folded paper receipt resting on the fabric floor nearby, enormous next to them Do not crop so tight that only fabric walls remain. If the pocket silhouette would leave the frame, pull the camera back. CONTINUOUS MOTION REQUIRED: Real, visible physical motion for almost the entire 8 seconds — a clear sequence of distinct actions, never one static pose held for seconds at a time. SCENE: The subject, shrunk to a few centimeters tall, is inside that jeans pocket. A coin and a folded receipt rest on the fabric floor nearby. Indoors, soft natural window light from one side so fabric texture and the subject both read clearly. The wearer of the jeans is a different full-sized adult — never give them the subject’s face. No gym, no banners, no second copy of the subject. 0-1.5s: Wide enough to read the jeans pocket on the hip. Tiny subject sits calmly inside the pocket bag, idly straightening the folded receipt beside them, at ease in their little space. Waistband and belt loop still in frame. 1.5-3s: A phone begins ringing in a nearby pocket — the entire fabric floor starts buzzing and shaking rhythmically beneath them. Startled, the subject grabs onto a fold of fabric to keep their balance as the coin and receipt rattle and skitter across the floor. The U-shaped pocket opening stays readable. 3-5s: The vibration intensifies. The subject stumbles and slides across the shaking fabric, arms wheeling, trying to stay upright as the whole pocket pouch trembles around them — a genuinely comedic struggle against the motion. Jeans waistband still in frame. 5-6.5s: The vibration cuts off abruptly. The subject, thrown off-rhythm, stumbles forward and catches themselves against the fabric wall, dizzy for a beat, blinking. Pocket silhouette still readable. 6.5-8s: They shake off the daze, laugh, straighten up, and peek upward through the pocket opening at the now-quiet world above, giving a small relieved thumbs-up toward camera. Waistband still readable. SCALE REALISM: The pocket’s real stitching, fabric weave and denim texture must read at true size next to the tiny subject — the coin, the receipt and every fold of fabric should look enormous relative to them. Match lighting and color grade between the subject and the fabric throughout so the whole shot reads as one real photograph, not a composite. Keep the camera framing wide enough that the shaking, the stumble and the recovery all stay clearly visible in frame. No on-screen text, no watermark, no subtitles.

2 steps

MODEL: Seedance 2.5 — use this model specifically. Do NOT use MiniMax or any other model for this generation. FORMAT: Vertical video, 9:16 aspect ratio. DURATION: 8 seconds. AUDIO: Generate synchronized natural audio for this scene — soft fabric rustling as the subject climbs and grips the pocket edge, a sharp metallic clink when the coin drops in, a startled gasp, and a genuine laugh at the end. No music, no voiceover, no narration. ASSET MAP: @image1 — the subject. SUBJECT LOCK — HIGHEST PRIORITY: Preserve the subject in @image1 exactly. Exact face, facial features, hairstyle, skin tone, build and gender presentation stay unchanged — only their scale changes. Do not beautify, restyle or alter identity. Photocopy @image1. Face, hair, skin, and the white collar/shirt from the still stay. POCKET READ — MUST BE OBVIOUS FROM FRAME 1: This is a real jeans FRONT POCKET on a pair of blue denim jeans worn at the hip of a different, full-sized adult (never the subject’s face). A first-time viewer must instantly recognise it as a clothing pocket — not a fabric cave, not a bag, not a tunnel of cloth. Camera sits OUTSIDE the garment, slightly above hip height, looking down and in through the pocket opening. In every frame keep visible at once: - the jeans: waistband, a belt loop, bartack stitching, rivet, denim weave - the pocket as a 3D pouch sewn onto the jeans, with a clear U-shaped opening and a folded pocket-edge hem - the tiny subject inside that pouch, a few centimetres tall, with the pocket bag huge around them Do not crop so tight that only fabric walls remain. If the pocket silhouette would leave the frame, pull the camera back. CONTINUOUS MOTION REQUIRED: Real, visible physical motion for almost the entire 8 seconds — distinct actions, never one static pose held for seconds. Indoors, soft natural window light from one side. The wearer of the jeans is a different full-sized adult — never give them the subject’s face. 0-1.5s: Wide enough to read the jeans pocket on the hip. Tiny subject sits inside the pocket bag, looking up at the opening and the light. Waistband and belt loop still in frame. 1.5-3.5s: They grip the folded pocket-edge hem with both hands and climb, hoisting themselves up until perched on the rim of the pocket opening, legs dangling over the outside of the jeans, catching their breath. The U-shaped pocket opening stays readable. 3.5-5s: The person wearing the jeans shifts their weight; the whole pocket pouch sways. The subject grips the hem tightly, body rocking, still clearly inside a jeans pocket on a hip. 5-6.5s: A coin drops in from above through the pocket opening and lands with a clatter beside them in the pouch — the subject flinches back, startled, both arms raised. The coin is enormous next to them. Pocket opening still visible. 6.5-8s: They relax, walk over and pick up the coin with both hands, straining against its size, then hold it up over their head in triumph, laughing — still standing inside the jeans pocket, waistband in frame. SCALE REALISM: Stitching, denim weave, belt loop and coin are true size next to the tiny subject. Match lighting so it reads as one photograph, not a composite. No on-screen text, no watermark, no subtitles.

2 steps

MOTION SOURCE — this recipe was reconstructed from a source clip hosted on this platform, which is the reference for the choreography: https://cdn.openclips.ai/community_workflows/3c31cff5-41fd-4c04-9704-df1df83a5657.mp4 (15s, 9:16, the original 'silly cat' dance — conformed to exact 9:16, audio intact) The dance beats written into step 2 were transcribed from that clip frame by frame at 2fps: the springy two-step, the alternating front paws, the knee lifts, the head bobs, the tail counterweight, and the wide-shrug and raised-paw accents. The one-legged spin is an addition, not in the source. IMPORTANT — the clip is NOT passed to the model, and cannot be. Two placements were tested and both fail: inside a step's params it is refused outright ('params carries videoUrl, which seedance-2.0 does not take'), and at step level it is accepted by the API and then silently dropped from the stored row. So step 2 DESCRIBES the choreography instead of copying it, and users get these moves at the model's own timing, not beat-for-beat with the source. If videoUrl is ever accepted on a step, attach the clip above as the motion reference on step 2 and the dance will be copied beat for beat — that is how the preview video on this workflow was actually produced (Seedance 2.0, videoUrl + character + backdrop reference, via a direct generate call rather than a workflow run). AUDIO: the source clip's own soundtrack cannot ride either — audio: 'original' needs Kling's videoUrl, and audioUrls is refused at render. Step 2 uses generateAudio instead, prompted for a silly bouncy beat.

2 steps

MODEL: Seedance 2.5 — use this model specifically. Do NOT use MiniMax or any other model for this generation. FORMAT: Vertical video, 9:16 aspect ratio. DURATION: 12 seconds. AUDIO: Generate synchronized natural audio for this scene — soft ambient fabric rustling, a held breath as the subject hides, a loud metallic jingle when the keys are pulled out, then a relieved exhale and a soft triumphant exclamation at the end. No music, no voiceover, no narration. ASSET MAP: @image1 — the subject. SUBJECT LOCK — HIGHEST PRIORITY: Preserve the subject in @image1 exactly. Exact face, facial features, hairstyle, skin tone, build and gender presentation stay unchanged — only their scale changes. Do not beautify, restyle or alter identity. Photocopy @image1. Face, hair, skin, and the olive-green quarter-zip over the white tee from the still stay. POCKET READ — MUST BE OBVIOUS FROM FRAME 1: This is a real jeans FRONT POCKET on blue denim jeans worn at the hip of a different, full-sized adult (never the subject’s face). A first-time viewer must instantly recognise it as a clothing pocket — not a fabric cave, not a bag, not a tunnel of cloth. Camera sits OUTSIDE the garment, slightly above hip height, looking down and in through the pocket opening. In every frame keep visible at once: - the jeans: waistband, a belt loop, bartack stitching, rivet, denim weave - the pocket as a 3D pouch sewn onto the jeans, with a clear U-shaped opening and a folded pocket-edge hem - the tiny subject inside that pouch, a few centimetres tall Do not crop so tight that only fabric walls remain. If the pocket silhouette would leave the frame, pull the camera back. CONTINUOUS MOTION REQUIRED: Real, visible physical motion for almost the entire 12 seconds — a clear sequence of distinct actions (settle, alert, hide, reaction, climb, look out), never one static pose held for seconds at a time. SCENE: The subject, shrunk to a few centimeters tall, is inside that jeans pocket. A folded receipt and a set of keys rest on the fabric floor nearby. The pocket opening is above them. Indoors, soft natural window light from one side. The wearer of the jeans is a different full-sized adult — never give them the subject’s face. 0-2s: Wide enough to read the jeans pocket on the hip. Tiny subject sits comfortably inside the pocket bag, arranging the folded receipt like a small mat beneath them, settled and at ease. Waistband and belt loop still in frame. 2-4s: Bright light suddenly widens above as the pocket flap is pulled open — they look up sharply, alert, rising to their feet. The U-shaped pocket opening stays readable. 4-6s: A pair of giant fingers (the wearer’s, not the subject) reach down into the pocket, searching side to side. The subject scrambles and presses themselves flat into a fold of fabric in the shadowed corner, perfectly still, visibly holding their breath. 6-8s: The fingers close around the keys resting just beside the subject’s hiding spot and lift them straight up and out of the pocket — the keys jingle loudly, a gust of motion ruffling the subject’s hair and clothing as they pass close by. Keys and fingers are enormous next to them. 8-10s: The fingers withdraw and the flap falls closed behind them, light dimming back down. The subject peeks out from the fold, relieved, then breaks into motion, climbing quickly up the inner fabric wall toward the pocket’s rim. Pocket silhouette still in frame. 10-12s: They reach the rim and perch on the edge, silhouetted against the bright opening, looking out at the towering room beyond, then turn back toward camera and give a small triumphant fist pump, grinning. Waistband still readable. SCALE REALISM: Stitching, denim weave, keys and fingers are true size next to the tiny subject. Match lighting so it reads as one photograph, not a composite. No on-screen text, no watermark, no subtitles.

2 steps

MODEL: Seedance 2.5 — use this model specifically. Do NOT use MiniMax or any other model for this generation. FORMAT: Vertical video, 9:16 aspect ratio. DURATION: 8 seconds. AUDIO: Generate synchronized natural audio for this scene — soft fabric rustling as the subject climbs and grips the pocket edge, a sharp metallic clink when the coin drops in, a startled gasp, and a genuine laugh at the end. No music, no voiceover, no narration. ASSET MAP: @image1 — the subject. SUBJECT LOCK — HIGHEST PRIORITY: Preserve the subject in @image1 exactly. Exact face, facial features, hairstyle, skin tone, build and gender presentation stay unchanged — only their scale changes. Do not beautify, restyle or alter identity. Photocopy @image1. Face, hair, skin, and the white collar/shirt from the still stay. POCKET READ — MUST BE OBVIOUS FROM FRAME 1: This is a real jeans FRONT POCKET on a pair of blue denim jeans worn at the hip of a different, full-sized adult (never the subject’s face). A first-time viewer must instantly recognise it as a clothing pocket — not a fabric cave, not a bag, not a tunnel of cloth. Camera sits OUTSIDE the garment, slightly above hip height, looking down and in through the pocket opening. In every frame keep visible at once: - the jeans: waistband, a belt loop, bartack stitching, rivet, denim weave - the pocket as a 3D pouch sewn onto the jeans, with a clear U-shaped opening and a folded pocket-edge hem - the tiny subject inside that pouch, a few centimetres tall, with the pocket bag huge around them Do not crop so tight that only fabric walls remain. If the pocket silhouette would leave the frame, pull the camera back. CONTINUOUS MOTION REQUIRED: Real, visible physical motion for almost the entire 8 seconds — distinct actions, never one static pose held for seconds. Indoors, soft natural window light from one side. The wearer of the jeans is a different full-sized adult — never give them the subject’s face. 0-1.5s: Wide enough to read the jeans pocket on the hip. Tiny subject sits inside the pocket bag, looking up at the opening and the light. Waistband and belt loop still in frame. 1.5-3.5s: They grip the folded pocket-edge hem with both hands and climb, hoisting themselves up until perched on the rim of the pocket opening, legs dangling over the outside of the jeans, catching their breath. The U-shaped pocket opening stays readable. 3.5-5s: The person wearing the jeans shifts their weight; the whole pocket pouch sways. The subject grips the hem tightly, body rocking, still clearly inside a jeans pocket on a hip. 5-6.5s: A coin drops in from above through the pocket opening and lands with a clatter beside them in the pouch — the subject flinches back, startled, both arms raised. The coin is enormous next to them. Pocket opening still visible. 6.5-8s: They relax, walk over and pick up the coin with both hands, straining against its size, then hold it up over their head in triumph, laughing — still standing inside the jeans pocket, waistband in frame. SCALE REALISM: Stitching, denim weave, belt loop and coin are true size next to the tiny subject. Match lighting so it reads as one photograph, not a composite. No on-screen text, no watermark, no subtitles.

2 steps