How to make this?

프로젝트 / 카피바라 · 2026년 9월 7일

카피바라

아침 주차장, 리모컨 한 번에 카피바라가 승합차만 해집니다. 여자는 올라타서 출근합니다. 아무도 놀라지 않습니다.

따라 만들기
Seedance 2.5모델명
30총 길이, 초
2클립
5References
1생성 결과

발행일 2026년 9월 7일

제작기

아침 주차장에 손바닥만 한 카피바라가 앉아 있고, 여자가 리모컨을 눌러 승합차 크기로 키운 다음 올라타고 회사로 갑니다. 이 작품의 엔진은 톤 하나입니다. 황당한 부분을 나머지 평범한 부분과 똑같이 평범하게 찍는 것. 그래서 거대화 장면에 빛나는 효과도 입자도 넣지 않았습니다. 대신 카피바라 밑에서 주차 스토퍼가 사라지고, 자갈이 링 모양으로 튀고, 옆에 세워둔 자전거가 밀린 공기에 넘어집니다. 크기가 변했다는 걸 빛이 아니라 물리로만 말하게 했습니다. 여자도 카피바라도 30초 내내 무표정이고, 마지막 인사 한 마디만 밝습니다. 레퍼런스는 5장입니다. 주인공 캐릭터 시트, 카피바라 크리처 시트, 리모컨 소품 시트, 로케이션 두 곳. 카피바라 크리처 시트에는 크기를 알려주는 단서를 일부러 하나도 넣지 않았습니다. 같은 동물이 영상 안에서 고양이만 하다가 승합차만 해지고 다시 작아지기 때문에, 시트가 크기를 정해버리면 연출과 싸웁니다. 크기는 전부 각 구간의 문장이 정합니다. 리모컨은 특수 장비로 그리지 않았습니다. 누렇게 변색된 낡은 가정용 리모컨에 규격에 없는 빨간 버튼 하나만 비뚤게 박아 넣고, 나머지는 눌린 자국과 닳은 인쇄, 떨어뜨린 모서리 흠집으로 채웠습니다. 물건 전체를 특별하게 만들면 웃음이 사라집니다. 빔은 두 번 나옵니다. 커질 때와 작아질 때. 두 구간 모두 버튼을 누르는 순간부터 빛이 닿고, 몸이 부풀고, 손을 떼면 잔상 없이 사라지는 과정을 각각 다시 적었습니다. 한 번만 써두면 뒷구간에서 효과가 그래픽처럼 미끄러지기 때문입니다. 좌우 배치도 전 구간 고정했습니다. 자판기는 화면 왼쪽, 담벼락은 오른쪽. 컷이 바뀌어도 뒤집히지 않게 잠가두면 시선이 어디를 향하는지가 흔들리지 않습니다.
단계

따라 만들기

각 단계의 프롬프트를 복사해 이미지를 만들고, 이미지마다 붙은 @Image 번호 순서 그대로 업로드하면 마지막 영상 프롬프트의 번호가 정확히 맞습니다.

Step 1

캐릭터 · 이미지 2장

주인공 여자 캐릭터 시트

1번째로 업로드하면 프롬프트의 @Image 1이 됩니다

Character reference sheet as FOUR full-body views in a row on a clean white studio background - FRONT, 3/4, SIDE PROFILE, BACK - plus one large face close-up inset panel at the right. The same woman in every view, same outfit, same relaxed neutral standing logic, arms hanging naturally at her sides, both hands completely empty.

Proportions: a tall lookbook-model figure, about nine heads tall, small head, long legs carrying well over half of her standing height, and in every view she stands head-to-toe filling the full height of the panel. Do not compress or shorten the figure to fit the frame.

Outfit, a full early-2000s look - a fitted pale lavender ribbed cotton baby tee, cropped to sit just above the navel, with cap sleeves and a tiny satin bow stitched at the centre of the scoop neckline; over it a cropped light-wash denim jacket worn open, boxy through the shoulders with contrast stitching, a raw hem and the sleeves pushed to the elbow; a low-rise light-wash denim mini skirt riding on the hip bones, with a front zip, belt loops and frayed hem, worn with a thin white leather belt closed by a small silver oval buckle; white chunky platform sneakers with flat laces and a ribbed rubber sole.
worn accessories: a short silver chain choker strung with small freshwater pearls sitting high on the throat, two small butterfly hair clips in frosted lilac pinning back the crown, tinted rose-lens rectangular sunglasses resting on top of the head, a stack of three beaded bracelets in lilac, silver and clear glass on the left wrist, a slim silver ring on the right hand, and a tiny quilted shoulder bag in cream faux leather on a short strap tucked under the arm, a small silver heart charm hanging from its zip.
Hair - dark brown with a warm cast, straight and very glossy, parted in the centre and falling just past the shoulders with blunt ends, the crown pinned back by the butterfly clips and two fine face-framing strands left loose along the cheeks.
Makeup - dewy skin with a pearly sheen on the high points, frosted pearl-pink shimmer shadow washed over the lids with a silver-white dot at the inner corner, long defined lashes, soft brown liner smudged close to the lash line, high pink blush across the cheeks and the bridge of the nose, and glossy pink-nude lips with a soft liner edge.

Identity strictly from the attached photo: her facial features, skin, hair colour and hair texture come from that image and nothing else. The attached photo supplies identity only - do not carry over its ballet studio background, its mirror and barre, its cream long-sleeve top and draped satin wrap skirt, its leg warmers, its severely slicked-back ponytail or its side-standing pose; every panel here is the neutral white-background turnaround described above, standing straight, lit as its own studio session.

Clean even studio lighting like a fashion lookbook turnaround, true fabric colour and garment weight, visible rib knit, denim twill and the sheen of the satin bow, real skin with pores and natural catchlights in the eyes. The face close-up inset is a straight-on frontal head-and-shoulders crop, large enough to read every feature clearly, with a calm neutral expression. Photorealistic, ultra-detailed, 4K. No text, no logo, no watermark.
카피바라 「모찌」 크리처 시트

2번째로 업로드하면 프롬프트의 @Image 2이 됩니다

Creature reference sheet as THREE panels side by side on a clean light-grey studio background (#c9c9c9) — PANEL 1 full-body SIDE profile standing, PANEL 2 full-body FRONT view sitting with front paws tucked, PANEL 3 large head close-up. Subject — one capybara with a broad blunt muzzle, small rounded ears set far back, tiny dark almond eyes with a permanently calm half-lidded expression, a heavy barrel body on short sturdy legs, and no visible tail. Coat is coarse wiry guard hair in warm sandy brown with a slightly darker chestnut saddle across the spine and paler oatmeal fur on the cheeks, throat and belly; the hair lies in uneven cowlicks and sticks out in tufts around the shoulders and rump, dry and a little dusty at the tips. Faint pink skin shows through the sparse fur on the muzzle bridge and around the nostrils, whiskers short and few, front paws with four blunt nails, hind feet slightly webbed. Shot like a real studio animal portrait session on a medium-format camera, soft even light with a gentle falloff behind the subject, true fur grain, natural moisture on the eyes and nose, a couple of loose hairs catching the light. Same individual animal in all three panels. Original design, not resembling any existing film or cartoon creature. Photorealistic, ultra-detailed, 4K. No text, no logo, no watermark.
Step 2

장소 · 이미지 2장

코인 주차장 로케이션

4번째로 업로드하면 프롬프트의 @Image 4이 됩니다

A location-scout photograph of a tiny Japanese coin-operated parking lot wedged between two low residential buildings in a quiet suburban back street, taken at a 3/4 angle from the street entrance on a Canon EOS 5D with a 35mm lens. Early morning, flat overcast daylight with one weak shaft of sun striking the far wall, no harsh shadows. Four numbered bays of cracked asphalt with faded white paint lines, low yellow flap-locks raised on the empty bays, concrete wheel stoppers stained with tyre rubber, a squat grey payment terminal at the entrance, a dented drinks vending machine humming against the left-hand wall, a row of blue plastic crates and a coiled hose beside it, weeds pushing through a split in the asphalt at the back corner, an aluminium bicycle leaning on the right-hand cinder-block wall under a tangle of overhead power lines. Lived-in detail everywhere — water stains bleeding down the block wall, a rusted drain grate, cigarette scorch marks near the terminal, mismatched patches of newer tarmac. Real architecture, physically grounded objects, true light falloff, subtle film grain, slightly imperfect framing like a real scouting snap. Empty of people and empty of cars. Any signage is small, distant and unreadable. NOT a 3D render, NOT CGI, NOT a matte painting, NOT AI-looking. No readable text, no logo, no watermark. Ultra-detailed, 4K.
회사 주차장 로케이션

5번째로 업로드하면 프롬프트의 @Image 5이 됩니다

A location-scout photograph of the open-air staff parking area behind a mid-sized Japanese office building, taken at a 3/4 angle from the far end of the lot on a Canon EOS 5D with a 35mm lens. Early morning, flat overcast daylight, long soft reflections on damp asphalt from rain the night before. A grey tiled office block six storeys high fills the background with rows of plain aluminium-framed windows and a glass entrance lobby up three shallow steps, a stainless handrail down the middle of the steps. Neat rows of numbered bays in fresh white paint, several compact sedans and a white van already parked at the far side, a green mesh fence along the boundary, a covered bicycle rack half full, a bank of grey utility boxes and a single ashtray stand near the lobby doors, a young planted tree in a square concrete bed. Lived-in detail — kerb paint scuffed by tyres, a stack of flattened cardboard by the service door, oil ghosting on two of the bays, a fire extinguisher cabinet with sun-bleached paint. Real architecture, physically grounded objects, true light falloff, subtle film grain, slightly imperfect framing like a real scouting snap. Empty of people. Any signage is small, distant and unreadable. NOT a 3D render, NOT CGI, NOT a matte painting, NOT AI-looking. No readable text, no logo, no watermark. Ultra-detailed, 4K.
Step 3

거대화 리모콘 소품 시트

거대화 리모콘 소품 시트

3번째로 업로드하면 프롬프트의 @Image 3이 됩니다

Product reference sheet of a small handheld remote control as FOUR views on a clean light-grey studio background — front face straight on, back showing the battery hatch, left side edge, and a three-quarter view — plus ONE in-use inset of a bare hand holding it at chest height with the thumb resting on the large red button. The remote is an old household air-conditioner type: a rounded rectangular shell in yellowed off-white plastic about the length of a palm, a narrow dark grey rubber keypad of small worn buttons whose printed markings have rubbed away to faint ghosts, a tiny blank LCD strip at the top, a black infrared lens window at the tip. ADDED and clearly aftermarket: one oversized glossy red arcade-style push button screwed into the upper shell, seated in a slightly crooked hand-cut hole with a hairline crack radiating from one screw and a smear of grey adhesive at the seam, a short length of red wire disappearing back inside. Honest wear everywhere — greasy sheen on the most-pressed keys, a hairline scratch across the LCD, dust packed into the seams, one corner scuffed white from being dropped. Shot like a commercial product photo set, soft even light, razor-sharp focus, true plastic sheen. Photorealistic, ultra-detailed, 4K. No readable text, no logo, no watermark.
Step 4

영상 프롬프트

본편
REFERENCE ROLES
@ {{Mixed 1}} — the WOMAN: face, hair, makeup, outfit. Ignore its white background, lookbook pose, studio light and framing; camera, posing, expression and light come from the shots below.
@ {{Mixed 2}} — MOCHI the capybara: body shape, fur colour and texture, face. Ignore its grey background and take no scale from it — his size is set per timecode below and changes twice.
@ {{Mixed 3}} — the REMOTE: shape, yellowed plastic, worn keypad, oversized red button. Ignore its background and inset hand.
@ {{Mixed 4}} — OPENING LOCATION, a coin parking lot: geometry, surfaces, walls, vending machine, wheel stoppers, daylight. Ignore camera and framing. Do not widen, add buildings or upgrade it.
@ {{Mixed 5}} — COMPANY PARKING LOCATION, same terms: bays, office block, lobby steps, daylight.

WARDROBE
Exactly the outfit in @ {{Mixed 1}} , no part of it ever changing: cropped denim jacket with the sleeves pushed to the elbows, lavender ribbed crop camisole, low-rise denim micro miniskirt with a white belt, chunky white platform sneakers, small cream shoulder bag at her right hip, pearl choker, pink beaded bracelet on her left wrist, sunglasses pushed up on her head, lavender butterfly clip. Long straight dark brown hair, centre parted, two strands in front. Pink blush high on her cheeks and under her eyes, glossy lips.

LANGUAGE AND VOICE
Japanese only, lips precisely matching the pronunciation. Light, low-energy, matter-of-fact — the way you talk to a pet you see every morning. Exactly three lines are spoken out loud and four are inner voice-over, in the order below and no others. Her mouth moves only on the out-loud lines; on every inner voice-over her lips stay closed and still. Mochi never speaks and his mouth never syncs to any sound.

STYLE
Handheld documentary realism, a Japanese morning commute vlog, 35mm mirrorless: overcast daylight, no grading beyond a cool cast, real skin texture with visible pores, blush reading as makeup not a filter, real denim weight and fray, subtle grain, micro-shake and a half-beat of focus hunting. The absurd part is shot exactly as plainly as the ordinary part — no fantasy glow, no epic angles.

SETTING AND OBJECT REGISTER
Two locations, nothing else. From the first frame, before she arrives, the coin lot holds the humming vending machine screen-LEFT, the cinder-block wall with the leaning bicycle screen-RIGHT, three raised yellow flap-locks and four concrete wheel stoppers across the middle ground. That left-right arrangement is locked in both locations; the camera never flips it. Cat-sized Mochi already sits on the middle stopper. She brings in only the bag, and the only thing that ever leaves it is the remote, out once and back once, right hand; her free left hand rests at her thigh unless doing something named below. Exactly one named character appears; other people are distant passers-by, out of focus, faces resembling no reference, never approaching, never reacting.

SOUND
Location sound only: the vending machine's hum, sparrows, a crossing bell, a bicycle on the next street. Her soles scuff asphalt; the remote clicks hard and plastic. The growth is a low pressure thump felt more than heard, then grit skittering and a car alarm two streets off. Full-size Mochi breathes slow and heavy, weight landing in soft flat pads, gut sloshing faintly each step. Riding, the street rises — traffic a block over, a shutter rolling up, a crosswalk chirping. At the company lot it drops to ventilation fans and a distant car door. No silent gaps.

ACTION, in timecode

0.0-3.5 — EMPTY LOT. Static wide from the street entrance, chest height. Nothing happens for two seconds: overcast light, weeds shifting. Cat-sized Mochi sits on the middle stopper — it is wider than he is — blinking slowly, facing screen-left. A sparrow lands on the far wall and leaves.

3.5-6.5 — SHE ENTERS. Same framing. She walks in from screen-LEFT past the vending machine, sneakers scuffing, bag swinging once, and stops three metres from the stopper facing screen-right. She looks down at him, he looks up at her, their eyeline crossing the frame. Her face does nothing at all — no smile, no greeting.
INNER V.O.: 「また、そこ。」

6.5-9.0 — THE REMOTE. Cut to a closer three-quarter of her from Mochi's side, just above stopper height, so she reads tall against the sky with his ears soft in the foreground screen-right. Her right hand goes into the bag and comes out with the remote. She turns it over once, thumbs the worn keypad out of habit, settles her thumb on the big red button, lifts it at him. Her left hand hangs at her thigh.
INNER V.O.: 「電池、まだいけるか。」

9.0-13.0 — THE BEAM AND THE GROWTH. HARD CUT to a wide two-shot from the street entrance: her on the LEFT facing screen-right, Mochi on the RIGHT on the stopper, three metres of asphalt between them. Her thumb presses. The remote clicks and a thin flat beam of pale white light leaves the lens window straight across the gap onto his shoulder — physical light with visible falloff and a faint haze in the morning air, alive only while her thumb is down. Under it Mochi expands smoothly over two and a half seconds with enormous mass: the stopper disappears under him, his weight settles onto the asphalt with a low thump, grit skitters outward in a ring, the flap-locks vanish beneath his belly, the bicycle topples against the wall from the displaced air. He ends van-sized, five metres nose to tail, his back a metre eighty up — her head now level with his shoulder, and she tilts her chin to keep looking at him screen-right. Her thumb lifts and the beam is gone, no residue. His expression has not changed at all. She lowers the remote and puts it back, same hand.

13.0-16.0 — MOUNTING. Cut low from his front-left, camera near the ground, her small against his flank. He folds his front legs and lowers his chest with the slow effort of a heavy animal, loose skin bunching at the shoulder. She steps onto his folded foreleg, grabs a handful of coarse guard hair, swings her right leg over and settles astride behind the shoulders, one hand on his neck, the bag pulled to her lap, her free hand smoothing the skirt hem as she lands.
OUT LOUD, flat and quiet, like talking to a bicycle: 「おはよ、モチ。」

16.0-19.5 — LEAVING. Cut to a following shot from behind and above as Mochi rises and walks out screen-left into the residential back street. Parked cars come up to his chest; he clears the low gate posts with no room to spare. His back rolls side to side so she rocks with it and puts a hand down to steady herself.
INNER V.O.: 「電車より、はやい。」

19.5-23.0 — THE STREET. HARD CUT to a lateral tracking shot from across a two-lane street, long lens so the traffic compresses behind them. Mochi walks the leftmost lane at the pace of traffic; a compact car passes him on the outside without slowing, a delivery scooter goes the other way. She rides looking ahead over his head, one hand in his shoulder fur, hair lifting.
OUT LOUD, quiet, aimed down at the back of his head, no urgency: 「モチ、次そこ右。」
His ear flicks once and he drifts right at the corner without breaking stride.

23.0-26.5 — ARRIVAL AND SHRINK. HARD CUT to a static wide of the company lot from its far end: office block and lobby steps screen-RIGHT, bicycle rack and fence screen-LEFT, locked for the rest of the film. Mochi walks in from screen-left, stops in an empty bay and lowers his chest again. She slides down off his flank, straightens her skirt and jacket, takes the remote out and steps two paces back, pointing it up at his shoulder. Click, thumb held down: the same flat white beam, same falloff and haze, and Mochi deflates smoothly over two seconds with the same enormous-mass logic in reverse — his shadow shrinking under him, the displaced grit settling back — until he is cat-sized again mid-bay. Her thumb lifts, the beam is gone. The remote goes back in the bag.
INNER V.O.: 「ここでは、ふつうの顔。」

26.5-30.0 — GOOD MORNING. Cut to a closer three-quarter of her from the lobby steps, low, the office doors soft behind her screen-right. She crouches, scoops cat-sized Mochi up and sets him on her right shoulder, where he settles against her neck and goes still, front paws hanging. She stands, adjusts the bag strap, walks toward the steps screen-right and, starting up them, says it toward the doors ahead of her.
OUT LOUD, brighter and louder than every other line, a normal office greeting: 「おはようございます!」
Her face returns to flat neutral the instant the line ends. She takes the last step, the glass door swings, and the film ends on the empty bottom step with the quiet lot behind.

ACTING
Deadpan throughout. She treats a van-sized capybara as a commute, not an event — the work happens in her eyes and nobody plays the emotion. She never smiles, not even on the greeting. Mochi is a calm heavy animal at both sizes: slow blinks, slow breathing, one ear flick, never expressive, never reacting to the beam.

PHYSICS
Full-size Mochi carries real mass: his pads spread on contact, his belly and loose skin swing a half-beat behind his frame, his back dips as she settles. His gait is a real heavy quadruped walk, front and hind legs alternating strictly, one foot always on the ground, each pad landing flat. Coarse guard hair moves in clumps, not as a sheet. Denim holds its creases, sleeves stay pushed up, the frayed hem swings stiffly. Her soles land flat and audibly, heel first, strict left-right alternation. The bag swings and settles, never passing through her body.

NEGATIVE
No subtitles, on-screen text, readable signage, timestamps, logo or watermark. No background music at any point. No sparkle, glow, particles, lens flare or aura around Mochi. No slow motion; growth and shrink play at normal speed. Nobody ever looks at, faces or addresses the camera. No second capybara, no other animal, no crowd, no bystander filming. No object appears, disappears, floats or morphs into another. No identity drift between cuts, no duplicated limbs, no extra hands, no ghosting.

다른 작품

더 디테일한 제작 패키지 구상 중

지금 공개된 프로젝트들보다 더 디테일한 유료 패키지를 구상하는 중입니다. 이메일을 남겨 주시면 출시 후 가장 먼저 알려드릴게요.

카피바라 · How to make this?