【REFERENCE】 @image1 is a completed 2 columns × 2 rows 4-panel fashion editorial board featuring the same single subject in 4 different outfits. Define top left as PANEL_01／LOOK_01, top right as PANEL_02／LOOK_02, bottom left as PANEL_03／LOOK_03, bottom right as PANEL_04／LOOK_04. @image1 is an absolute lock for appearance: take face, hair, age appearance, skin tone, body type and proportions from it exactly and hold them identical across all LOOKs, with no drift or reinterpretation in any SHOT. Inherit clothing, shoes, accessories, background composition, base colours, accent colours and text only within the corresponding LOOK for each PANEL. Poses and shooting angles within PANELs are visual references from the start and not fixed. Do not newly set characters, outfits or colours on the text side; do not mix faces, bodies, outfits or colours between PANELs. Usage order is PANEL_01→02→03→04→01→02→03→04. Do not output 2×2 boards, borders, multiple people, or pans/zooms on sheets into the video; expand the corresponding PANEL into an independent full-frame space for each SHOT. 【CONDITION】 All 8 SHOTs, high-quality live-action fashion editorial video. Give all 8 SHOTs equal screen time; no SHOT runs noticeably longer or shorter than any other. Do not make it a video switching between 8 still posters. Ensure each SHOT establishes "starting pose → one clear fashion action → ending pose different from start," with stillness held only at the very end of the final SHOT. Combine natural one or half-step, body and shoulder direction turns, arm and hand position changes, weight shifts, head and gaze, and hair and outfit sway for the subject. Not dancing. Move subject, camera, text and colour blocks simultaneously. Base the entire edit on horizontal panel movement, without vertical scrolling. Within each SHOT, also move text, colour blocks, rectangles and lines left and right at 8–20% differing speeds. For each transition, shift the old LOOK's graphic plane 40–55% of screen width to one side, leaving about half, enter the next LOOK's colour block and text horizontally from the opposite side, pass through a brief uneven split screen, then fully reveal the next LOOK. Vary direction, width and speed per transition; do not repeat the same full-screen wipe. Only one subject during splits. Hard cut subject and outfit at edit points; do not use morphs or crossfades. Displayed text is only FASHION, IN MOTION, LOOK 01, LOOK 02, LOOK 03, LOOK 04, 02, 03. Same Helvetica Neue Condensed family as the adopted board. Make major text fully readable once before moving, without covering the face. 【SHOT FLOW】 [Shot 1] PANEL_01. Full body slightly to the right. From reference pose, move half a step, shift weight and arm position, stop in a different pose facing shoulder and gaze to camera. Camera short horizontal move opposite the subject. After FASHION and LOOK 01 establish, text and rectangle shift half screen left, LOOK_02 colour block enters from right. [Shot 2] PANEL_02. Full body with margins. While stepping horizontally, swing body to 3/4, change hand position to face front. Arc, base colour block and thin lines move horizontally at speeds different from the subject. At end, exit these background elements horizontally, enter PANEL_03 slope horizontally from opposite side. [Shot 3] PANEL_03. From start, the subject's full body fits centre-right on screen and does not follow the background slide. Without walking, from front-facing pose, relax one knee to shift weight to the hip, place one hand on the hip, return outward gaze to camera. Hair and hem sway naturally once with delay. Prohibit off-screen entry, horizontal move, sharp turns. Short push in. 03, slope, halftone and barcode accelerate horizontally, barcode strip slides horizontally to open LOOK_04. [Shot 4] PANEL_04. Full body. Step forward to open the shoulders, land in a pose with changed arm and leg placement. Short pull back. Arc, light colour block and IN MOTION shift left and right, at end retreat half to screen right, restart LOOK_01 grid from left. [Shot 5] PANEL_01. 3/4 to medium. During short push in, twist the upper body, quickly change one arm position and gaze to make the pose different from the first cycle. FASHION, rectangle and grid move horizontally in reverse direction parallax, connect to LOOK_02 rectangle. [Shot 6] PANEL_02. Pull back from medium to full body. Step forward to camera, change shoulder and hand position to land in a different full body pose. Base colour block, arc and thin lines three-layer horizontal move at differing speeds, at end exit these background elements horizontally, enter PANEL_03 slope horizontally from opposite side. [Shot 7] PANEL_03. From start, the subject is in screen centre and does not follow the background slide. During pull from knee-up to full body, slowly rotate about 90 degrees on one foot axis in place, turn the body then face front. Naturally lower both arms to stop in an upright pose. The long hem alone draws one large arc with delay. Maximise sway amplitude, do not increase subject speed. Prohibit off-screen entry, walking, fast horizontal move. Camera also short horizontal move. 03, slope, halftone and barcode cross left and right, open arc plane horizontally to connect to LOOK_04. [Shot 8] PANEL_04. Slowly pull from medium to full body. The subject steps forward to change arm, leg and weight, converging to a hero pose different from the first cycle. Arc and thin lines decelerate horizontal move, hold the end on a fully readable IN MOTION end screen. Do not return to 4 panels. 【NEGATIVE】 Prohibit sequence of still images, long stills per SHOT, repetition of reference poses, motion only at transitions, vertical scrolling, 2×2 board display, multiple people, subject duplication, pans/zooms on sheets, reading order changes, added people/outfits/colours/text, mixing between PANELs, different people, morphs of face/outfit/body, crossfades, all SHOTs same central full body or same scale, same-direction full-screen wipes, text cutoff, text covering face, subtitles, brands, logos, watermarks, limb distortions. overall_soundscape: Very soft, brief clicks and fabric swishes occur only at shot changes, mixed clearly below the music. No dialogue, narration or system voice. non_diegetic_music: A continuous instrumental fashion beat led by a dry kick drum, tight closed hi-hats and a steady synth-bass pulse at 118 BPM. It starts at a clearly audible level on the first frame, remains the dominant audio layer without ducking or fading, and continues through the final frame. No vocals.

4 steps63434.7K

30 seconds, 16:9, thirteen shots, upbeat music. @image_1 is the traveller and the only attached asset - the same person in every shot. ATTACHED IMAGES ARE MANDATORY AND FINAL: @image_1 decides the face, hair, skin and build in every frame. Where this text and the attached image disagree, the attached image wins. Preserve the gender presentation of @image_1 exactly as shown. Every hand in frame has five fingers and contact with a prop stays clean. WARDROBE LOCK: one outfit for all 30 seconds - a fitted indigo denim top, matching indigo denim bottoms, a burgundy leather belt and pointed burgundy ankle boots. Each garment is fixed; only its drape adapts to the body of @image_1. CAMERA: a phone travel video. @image_1 films the selfies at arm's length; an off-screen companion films the food handoff, the market walk from behind and the restaurant. Both operators stay behind the lens. 26mm-equivalent wide angle, colour as the phone records it, handheld movement on real footsteps, exposure and focus a beat late. Only the three restaurant shots from 17 to 23 seconds carry a DV-tape look - a texture over the whole picture, with the frame edge unchanged. [0-2s] Guesthouse window at morning, selfie already in motion, tilt up to a green-glass tower over low grey rooftops. "okay, Taipei, let's go." [2-4s] Main-road junction, scooters pulling off the stop line behind @image_1. "the scooters here are insane." [4-6s] Bubble-tea counter: the clerk lowers a clear sealed cup into the outstretched hand of @image_1 and lets go only once it is supported. [6-8s] Street stall: @image_1 puts both hands out and the vendor lowers a steaming white paper bag into them. [8-10s] Temple courtyard, tilt up off the face to incense smoke, red lanterns and a dragon-carved roof. [10-13s] Stepped hillside lane, tilt down to layers of red lanterns over a stone stairway in light fog. "this looks like a movie set." [13-15s] Dried-goods market, followed from behind at waist height, a bite taken while walking. [15-17s] Gondola with a glass floor, filmed straight down over green tea terraces. The music drops one beat. [17-19s] Restaurant, DV texture begins: a waiter lifts the lid off a bamboo steamer. "everyone says these explode." [19-21s] Macro: chopsticks open a dumpling and broth runs into the spoon. [21-23s] First mouthful, close and handheld. "okay. that's the one." [23-26s] Sunset waterfront, phone texture returns, focus racks from the bridge behind onto the face of @image_1. 60% slow motion. [26-30s] Night: a lit red sky lantern leaves the fingertips of @image_1, who waves at the lens. The hand covers the lens at 30 seconds. The thirteen shots run back to back with no gaps; the first frame is already lit and already moving. Writing in frame is only writing that belongs to the place. No subtitles.

2 steps36529.5K

@image1 is an absolute character lock for the figure: face, body, hair, outfit, materials and colour are taken exactly from it and must not drift or be reinterpreted at any point. @image2 is the graphic system and holds its off-white ground, mid-grey panel, sage teal and dusty rose blocks, thin black rules, corner tick brackets and type style throughout. @image3 is the storyboard and is followed exactly — same beat order, same framing, same staging, same scale. Style: a high-end editorial fashion lookbook. Photorealistic live-action figure, physically lit against a clean studio seamless, composited inside a flat Swiss-editorial graphic layout that rebuilds itself around them. Real skin, real denim, real jersey, fine fabric texture, natural depth of field, calm neutral colour, no heavy grade. The graphics stay flat, hard-edged and print-clean with no shadow or bevel. Movement is unhurried and confident throughout, never posed or bouncy, the whole piece reading as one continuous studio take with the layout changing around a figure who never cuts away. The figure walks toward camera in a relaxed three-quarter turn, one leg swinging through, arms loose at the sides. They continue the turn away from camera, then look back over one shoulder, the hips staying angled off. They come back around and walk straight toward camera, square to lens, arms swinging naturally with the stride. They arrive and settle, bringing one hand to rest on the hip as the weight drops onto one leg. They hold still and square with the arms at the sides, standing smaller in the frame with more ground around them. The camera moves closer to frame them from mid-thigh up, and they stand relaxed with the shoulders level. They turn slightly off axis at that same closeness, the weight shifting across, one shoulder easing toward camera. The camera draws back to full length and they finish with one hand on the hip, chin level, holding the pose. Through all of it the graphic layout rebuilds itself around the figure between beats: the masthead FASHION holds across the top, the labels LOOK_01, LOOK_02 and LOOK_03 exchange positions at the left and upper right, and the teal and rose blocks, thin rules and corner brackets slide and resettle into a new arrangement each time. The type is crisp, correctly formed and stable while on screen, never garbled, never flickering, never warping. （A calm minimal electronic music bed, soft and even, no vocals.） The figure is never duplicated and no second person enters. Do not generate any text beyond FASHION, LOOK_01, LOOK_02 and LOOK_03. Do not generate a logo. Do not generate a watermark. No filming equipment visible in frame. Skin and movement look like real footage, not AI-generated imagery, with no distorted hands, no extra fingers and no smeared contact where a hand meets the body.

5 steps1807.1K

Generate a UGC video, reviewing the product you want.

3 steps12613.1K

A locked static camera, eye level, subject centred and framed head to feet with the floor visible. The camera never moves, zooms or reframes. The setting is one modern apartment living room, unchanged from first frame to last: warm honey-toned wide-plank wood floor, cream-white walls, a low beige sofa with olive-green cushions along the back wall, a squared wood-slab coffee table holding a small stack of books, framed artwork on the left wall, tall black open shelving on the right with a trailing green plant, an open doorway at the right edge, soft even daylight. Along the bottom of the frame sits a row of four small circular white buttons, each holding a tiny icon of one garment. A small white arrow cursor moves between them. They stand centred in their own clothing, weight shifting from one leg to the other, hands settling at the waist, shoulders turning slightly. The cursor slides to the first button and taps it, and the outfit from @image1 replaces what they are wearing in a single instant snap. They smooth the skirt down and lift one hand to the brim of the hat. The cursor taps the second button and the outfit from @image2 snaps into place. They straighten, roll one shoulder to settle the pauldron, and take a half step. The cursor taps the third button and the outfit from @image3 snaps into place. They turn a few degrees to one side, set a hand on the hip, and let the weight shift open the slit. The cursor taps the fourth button and the outfit from @image4 snaps into place. They raise both hands to the back of their head, lower them again, and settle facing camera. Every outfit change is instantaneous, one frame to the next. No dissolve, no morphing, no transition, no fabric growing or unfolding. Room tone only. No dialogue, no speech, no singing, no music. Avoid generating any text, subtitles, captions, logos or watermarks.

6 steps13528K

ASSET MAP: @image_1 is the first frame of this film and the only reference. The person in @image_1 is the walker and the autumn village in @image_1 is the only location. Identity: the walker in every second is exactly the person in @image_1 - face, skin tone, hairline, hair length and colour, build and gender presentation unchanged, never masculinized or feminized, never restyled, and long hair is never added to a short haircut; the hair moves only as much as its own length allows. No action, gesture, camera note or lighting note anywhere in this text may imply hair that the person does not have: nothing is ever pushed, tucked, swept or blown out of the face, nothing is rimmed or backlit through hair, and where the hair is short or shaved it simply stays short or shaved and barely moves. Only one person walks in this film. The oatmeal cable-knit sweater, the light blue wide-leg jeans, the cream ankle boots, the timber houses, the cobbled lane, the bamboo fence, the firewood, the pumpkins and the amber hillside all come from @image_1 and stay unchanged. THE FORM - THIS IS THE WHOLE FILM: one single unbroken take. No cuts, no dissolves, no jumps and no change of location at any point. The camera never stops moving and never lets the walker out of frame; every change below is a continuous move of one camera, not an edit. Style: ultra-photorealistic live-action, warm late-autumn golden hour - low sun down the lane, amber and rust bounce from the orange canopy, long soft shadows on the cobbles, haze with drifting motes, fallen leaves lifting as the walker passes, real handheld weight, shallow depth of field. 0-5 seconds: the film starts exactly on @image_1 and moves at once - the walker lets go of the doorframe, steps down onto the cobbles and looks into the lens with a small easy smile before walking forward, the camera retreating ahead at walking pace, leaves drifting through the sunlight. 5-10 seconds: still retreating, the camera drifts lower and closer until the walker is held from the waist up, the low sun flaring past one shoulder while the orange canopy slides by on both sides. 10-15 seconds: without any cut the camera arcs around the shoulder of the walker and settles behind, following a few steps back down the lane past dark timber facades, a bamboo fence and drifts of fallen leaves. 15-20 seconds: still one move, the camera rises and pulls back into a high wide crane - the lane falling away, the whole village opening out with its roofs and woodsmoke, the amber mountainside behind and the walker small and centred far below. The move slows and the frame settles on the final second. Continuity: the walker always moves away down the same descending lane so the door is never seen again, five fingers on every hand, the light and the sun direction consistent for the whole take, and no other person ever in frame. Every surface stays free of lettering and signage. No green summer foliage anywhere; the whole film is deep autumn. Audio: boots on stone, dry leaves, light wind, a stream below, distant birds and a soft piano line. Nobody speaks. No subtitles.

3 steps1684.2K

THIRD-PERSON ACTION GAME CAPTURE. ONE PLAYABLE CHARACTER. NO DIALOGUE. FIFTEEN SECONDS. ASSET MAP - every reference is bound here and cited by tag below: @avatar_1 - the playable character, and the only named performer in the film. Face, hair, skin, age, build and every physical characteristic come entirely from @avatar_1 and never from this text. IDENTITY LOCK - HIGHEST PRIORITY: the playable character is exactly the person in @avatar_1. Keep their face, hair, skin tone and build the same from the first frame to the last. Do not restyle them, do not age them up or down, do not slim or bulk the build, and never swap in a different character. One playable character, the same one, for the whole capture. STYLISED, NOT PHOTOGRAPHED: @avatar_1 is rebuilt as a modern game character - clean stylised 3D shading, soft subsurface skin, rim light on the hair, crisp material separation - and never composited in as a photograph. Their face still has to read as unmistakably theirs: the same features, the same proportions, the same hair and skin tone, rebuilt in the game's own look. ATTACHED REFERENCES ARE MANDATORY AND FINAL: if a reference is attached, that is the character - reproduce them exactly, in every shot, and never invent a different one. The attached reference outranks every word written below it. FACE TIME: the camera sits behind and slightly above the right shoulder for most of the capture, but @avatar_1's face is clearly visible and clearly lit at least twice - once early, once at the end - as the character turns through the action. WARDROBE LOCK: @avatar_1 wears a fitted white and pale pink armoured jacket with hard shoulder plates, a utility harness across the chest, dark leggings with armoured shin guards, and a large rounded backpack. The backpack is white with warm yellow panels and it stays white and warm yellow in every single shot - it never turns green, blue, pink or any other colour. The whole outfit is fixed in cut and colour for the capture and only its drape adapts to the body wearing it. No lettering, no numerals and no logo anywhere on the clothing or the pack. THE LOOK: a captured frame from a modern third-person action game running at high settings. Clean stylised 3D rendering, physically lit, shallow game-camera depth of field, mild motion blur on fast turns, crisp specular highlights on metal and wet floor. Saturated colour, heavy use of practical light sources in frame. Nothing hand-drawn, nothing photoreal - it should read as a game, not as film. THE PLACE: a sealed underground facility. It opens in a cold teal service corridor with a wall-mounted data terminal glowing on one side. It moves into a red-lit loading corridor with heavy blast doors, warning chevrons painted on the metal, and pipework overhead. It ends in a long amusement arcade - two facing rows of upright arcade game machines with big lit marquees and glowing screens, the floor wet and reflecting magenta and cyan. THE OPPOSITION: three armoured figures in matte dark grey combat gear with full face visors. Faces never visible. No badges, no flags, no unit markings, no readable lettering anywhere on them. 00:00-00:03 Over the right shoulder of @avatar_1 in the teal service corridor, walking up to the wall terminal and reaching out to work it. The light from the screen falls across the side of the face. The camera holds behind the shoulder. 00:03-00:06 Cut to the red loading corridor: one of the opposition rushes in from the left and @avatar_1 turns hard into them, striking the weapon aside and putting them down in two fast movements. The face comes round to camera through the turn. The camera whips with the turn. 00:06-00:09 Cut to a low wide shot as a second figure comes over a crate and @avatar_1 drops them with a short burst, the muzzle flash lighting the blast doors, smoke rolling out across the floor. The camera tracks right at speed. 00:09-00:12 Cut to the arcade, camera behind the shoulder, @avatar_1 advancing slowly between the two rows of arcade machines with the weapon up. Marquee light slides across the armour plates. The camera pushes forward with them. 00:12-00:15 The last of the opposition breaks from cover at the far end and @avatar_1 drops them, then turns back toward camera, lit clean by the arcade screens, and holds. The camera settles to a stop and holds to the last frame. AUDIO: a sealed underground room - low air handling, a fluorescent tube buzzing, footsteps on wet concrete. Terminal beeps at the start, cloth and armour under strain, two short weapon bursts, shell casings on the floor, arcade machines idling in the background at the end. A tense electronic bed on low synth pulse, kept well under the effects and never rising to a peak. No voices, no shouting, no dialogue, no lyrics. Do not generate subtitles, captions, on-screen text, a mission banner, a minimap, a health bar, a HUD of any kind, a title card, a logo, a wordmark, a watermark or any readable lettering or numerals anywhere in frame. The frame must be left completely clean of interface. Do not generate spoken dialogue or narration.

2 steps13363

10-second 9:16 colorful giant-product city film. @image1 is the product: photocopy that tall black matte tumbler with handle, black lid, stainless rim and base. Not a soda can. No invented label. No on-screen text. 0.0–1.33s: Extreme macro of the matte black wall and handle in bright sunlight. Cut wider into cyan sky. 1.33–2.67s: The same tumbler floats among chunky pink and yellow voxel blocks and miniature cargo stacks. Geometric glyphs orbit outside the product. Energetic scale shift. 2.67–4.0s: Dramatic low city angle. Giant tumbler towers over stacked cargo. Hard sunlight, cyan sky. Full geometry intact, handle visible. 4.0–5.33s: Top-down miniature cargo yard. Oversized tumbler among pink, yellow, cyan stacks. 5.33–8.0s: Open-sky. Blocks assemble around the tumbler. One brief water splash ~6.4–7.2s without deforming it. Splash settles by 8.0s. 8.0–10.0s: Low-angle oversized hero against cyan sky. Tumbler fully visible, larger than surrounding structures. Hold. No captions. AUDIO: original electronic beat, dry clicks, one splash accent, no vocals.

2 steps066