Five steps, one generation each. Step 1 pick any avatar (@avatar). Step 2 turn the visitor's product photo (@product) into a three-view product map on grey so every later step sees the same details. Step 3 a two-frame reference sheet that locks the face, the hair, the jacket and the chained product together (Nano Banana Pro holds the face). Step 4 an eight-panel storyboard on Seedream to check the beats before paying for the 30s. Step 5 one unbroken 30s 16:9 Seedance 2.5 render, reference mode, with the sheet and the product map attached. The prompts never name the product or a gender, they read both from the images, so the recipe works with any avatar and any product the courier can hold in their hands. Known traps: never write 'a held beat of silence', the model renders it as a literal frozen frame, write tension as micro-motion instead; never write 'walks to the doors', the model walks to the END of the car, so stage the fight to finish beside the side doors; never cut to an outside camera on a door that already opened, the model closes and reopens it.
Five steps, one render. Steps 1 and 2 are your two avatars: Avatar 1 brings the gift and ends up running, Avatar 2 gets the gift and becomes the tiger. Any gender, any pairing, the prompts never say he or she. Step 3 builds one two-frame character sheet of both avatars in the film's wardrobe (Nano Banana Pro): this is where faces get locked and where you may recolor hair or restyle, never in the video prompt. Step 4 is one clean photo of your product, any product. Step 5 is the 30s video: @image1 is the sheet, @image2 is the product, all beats timestamped, the twist at 19s, the only camera move is the push-in on the product at the end. Fire as ONE 30s 9:16 Seedance 2.5 render with audio, 720p. Two traps: a small product is tiny in a wide shot, so the final beat pushes in on it while the tiger stays seated; and never let the product end up on the floor, that is where Avatar 1's shoe goes.
@avatar is the same real person in every shot — same face, same bone structure, filmed on location. Real skin with visible pores, individual hair strands, correct hands, natural expressions and anatomy. @avatar wears white wired earphones throughout, the cable bouncing with every movement. 0–4s CITY HOOK. Rain-wet street, cool daylight. 24mm, inches from the lens. @avatar gives a playful knowing smile, then turns and runs. Handheld chase. A red bus crosses extremely close and covers the frame. 4–8s OLD MARKET. The same stride continues into an atmospheric old town. Cream jacket, olive loose trousers. Real pedestrians, bicycles, stone, warm shopfronts. @avatar looks back laughing, then jumps a puddle. The camera dives toward the splash until water fills the frame. 8–12s MOUNTAIN REVEAL. The splash becomes the reflective floor of a glass lift. @avatar steps in, the doors close and reopen onto an enormous valley above the clouds. @avatar smiles back and runs out. Cold mist, huge depth, real wind. 12–16s COAST. The stride carries onto a golden-hour coastal road, ocean and cliffside architecture beyond. Deep-red track jacket, denim. The camera tracks alongside. @avatar spins, laughs, pulls the jacket off while still moving and throws it into the lens until red fabric fills the frame. 16–20s NEON NIGHT. The fabric clears into a rain-soaked neon city. Black leather jacket. Deep blacks, cyan reflections, red practicals, steam, real crowds. @avatar walks straight at the camera. A passing tram blocks the frame. 20–23s ABOVE THE CLOUDS. The tram windows become a cable-car window. Navy jacket. @avatar stands against the glass over mist-covered peaks, its reflection laid over the landscape. @avatar catches the camera in the reflection, smiles, and leaves frame. The camera whip-pans after it. 23–26s LOST. The whip resolves into a lively blue-hour plaza. The camera searches through real moving crowds. @avatar appears, disappears behind pedestrians, gone. Then a tap on the camera from behind — a fast whip around reveals @avatar standing there with a friend @avatar2, both laughing, and they run. 26–30s HERO ENDING. The camera chases them across the plaza. They hold hands, run, spin, laugh. Pigeons lift ahead of them. They turn back and throw their arms up. @avatar runs straight at the camera, laughing, and covers the lens with its hand. Hard cut to black. LOOK. Feature-film cinematography — 24mm chases, 35mm movement, 50mm portraits, handheld micro-movement, 180-degree shutter, real motion blur. Fine 35mm grain, subtle halation, soft bloom, expensive highlight roll-off. Colour runs cool rainy city, warm market, cyan mountain, golden coast, cyan-and-red neon, cool cable car, rich blue-hour finale. Skin stays naturally warm throughout. TRANSITION RULE. Every transition is physically motivated: bus wipe, splash, lift doors, jacket wipe, tram wipe, reflection whip, crowd reveal, hand to lens. Locations never visibly morph. No portals, no liquid transforms, no speed ramps, no floating camera. The cuts should be clean enough that a viewer replays the film to work out how they happened. REALISM. Weighted hair and clothing, real footsteps, physically correct reflections and splashes, the earphone cable always bouncing. Background people move independently — never frozen, never duplicated. No warped architecture, no plastic skin, no beauty filter. SOUND. Footsteps, traffic, the splash, a lift ding, wind, tram rumble, cable-car hum, crowd ambience, fabric, pigeon wings, real laughter. Music is a driving 120 BPM cinematic-electronic instrumental: minimal at the open, building through the chase, a melodic hook at the coast, deeper at neon night, stripped back for the cable car and the search, hard drop on the friend reveal, strongest for the final run. No voiceover, no lyrics.
Two steps, one generation each, and the product is a free variable. Step 1 takes the visitor's product photo (@product) and renders one eight-panel storyboard sheet on Seedream, one panel per scene, so the shot list and the product lock are checked for 19 tokens before the video is paid for. Step 2 is one MiniMax H3 15s 9:16 render at 2K in reference mode with the product photo and the storyboard attached. Every prompt describes the product only as 'the exact same product from the reference image' and says 'light accents in the product's own dominant colour', so the same text works for a watch, a bottle, a chocolate bar or a sneaker. Hands: at most two hands of one person, one calm pick-up and one touch, never a fastening, never a third hand. The beats that depend on the product type carry fallbacks inline (display waking, glass or liquid catching light, a lens flaring, or a light sweep across the logo). Rules that held: never ask for on-screen text, keep any screen or label identical everywhere, and if a Seedream job sits at processing with no progress for 10 minutes, re-fire it instead of waiting.
A sensual high-fashion film follows aa @avatar flowing through ten precise poses. The film unfolds in one uninterrupted 15-second shot inside a minimal white cyclorama with hard strobes, faint haze, glossy floor reflections, a satin sheet and scattered proof sheets. The camera follows one physically continuous path, accelerating between poses and braking sharply on every lock. 0:00-0:01.5 Pose 1: Centred frontal medium-close-up. @avatar faces forward with one hand gently brushing the hair beside their ear as the camera makes a restrained push-in. 0:01.5-0:03.0 Poses 2-3: @avatar turns into a three-quarter stance and subtly lifts their chin. @avatar then lowers their gaze as both hands settle softly near their collar and hair. The camera descends in a fast controlled arc, then pauses precisely on their eyes. 0:03.0-0:04.5 Poses 4-5: @avatar rotates into a strict left profile. @avatar rolls one shoulder forward while the edge of their dress catches a sharp side light. The camera whips smoothly past @avatar’s cheek and settles into a close pass. 0:04.5-0:06.0 Poses 6-7: @avatar gathers a fold of the clothes at their waist, then releases it and turns their face toward the lens in an over-the-shoulder look. The camera dives naturally to torso level and rises into a close facial pass. 0:06.0-0:07.5 Poses 8-9: @avatar lowers themselves onto the satin sheet with one knee raised. @avatar then extends one bare foot toward the lens, creating strong natural foreground foreshortening. The camera rockets forward at floor level and briefly holds on the perspective. 0:07.5-0:09.0 Pose 10: @avatar rises into a three-quarter stance with one hand resting near @avatar’s collarbone and the other touching their hair as the dress naturally skims along their leg. The camera slides rapidly across @avatar’s waistline before easing into a portrait hold. 0:09-0:11.0: Without repeating any previous pose, @avatar folds inward and closes their eyes for one controlled breath, then opens into a long upward stretch. The camera circles in a tight continuous orbit, moving slowly during the pose locks and rapidly through transitions. 0:11.0-0:13.0: @avatar twists into a rear three-quarter shoulder silhouette, then turns just enough for @avatar’s jawline and the deep-V neckline of the dress to catch the same sharp flash. The camera continuously skims from shoulder to face with pronounced natural parallax. 0:13.0-0:15.0: The camera grazes past @avatar’s eyes, follows the deep-V neckline, then arcs outward and lowers as @avatar settles into a confident dominant full-body pose, looking down into the lens. End on a clean wide hold beneath a final white flash bloom. VISUAL DIRECTION: Photorealistic human fashion-film aesthetic, realistic skin texture, natural facial details, physically accurate anatomy, realistic hair strands and movement, authentic fabric behaviour, cinematic studio lighting, glossy floor reflections, subtle atmospheric haze, high-end editorial fashion photography, sharp hard frontal flashes, defined side rim lighting, controlled exposure, realistic depth of field.
SUBJECT. The same @avatar throughout — 25, long honey-blonde hair with sun-lifted ends in a low messy bun with loose strands around their face, green-hazel eyes, freckles across their nose and cheeks, warm sun-kissed skin with real texture, minimal makeup, slim. Soft-yellow sleeveless knit top, loose light-grey trousers, white retro sneakers, dark-green shoulder bag, silver bracelet. Keep face, hair, outfit and accessories identical in every shot. 0–5s. @avatar leaves home with a shopping bag, locks the door, checks their keys, tucks a strand back into their bun, smiles at the camera, walks off down the street. 5–10s. @avatar buys a cup of cold cut mango with lime at the corner store window. Takes a bite, reacts to the sour lime, laughs, offers the cup toward the camera. 10–15s. A small dog behind a chain-link fence notices @avatar and wags. @avatar crouches to let it sniff their hand through the gap. The dog trots along the fence beside them, then turns back. 15–20s. Dry leaves and a few bougainvillea petals drop from the banyan onto their shoulder and hair. @avatar looks up and brushes the leaves away. The camera zooms in awkwardly. @avatar says, "Again?" 20–25s. @avatar sits on a low garden wall and finishes the fruit while a bicycle passes and a neighbour waters plants with a hose. Warm breeze in their hair and clothes; @avatar wipes their hands and adjusts their bag. 25–30s. @avatar walks toward the end of the street, turns back with a playful smile and says, "Let's go." Walks around the corner; the camera follows; abrupt cut to black. AUDIO. Location sound only — cicadas, birds, footsteps on cracked pavement, a bicycle bell, distant scooters, leaves, the hose, the dog, neighbourhood voices, a TV through an open window, camera-handling noise. No music, no narration, no captions.
FRAMING RULE: No full-body shots and no wide shots anywhere in the film. Only chest-up mid-shots, close-ups, extreme close-ups, shot-reverse-shot, and brief partial close-ups of hands, tea cup and ring. CONSISTENCY: Both keep identical faces, ages, clothing, hairstyles and positions throughout — no facial drift, no clothing change, no age variation. The @avatar1 is on the LEFT of the space, the @avatar2 on the RIGHT, separated by a small round table, facing each other or leaning slightly forward. Hold the 180-degree axis; eye lines match precisely in every shot. SHOT ONE (0.0-4.5s): 85mm portrait lens. @avatar1 in chest-up close-up, three-quarter profile facing screen right. Background is blurred ink-green curtain and rainy-night glass reflection; warm lamplight cuts across @avatar1’s face from the side. Camera nearly static with only the faintest handheld breathing. @avatar1 glances down at the ring box on the table, then lifts its eyes to @avatar2. Its brow does not visibly furrow but its gaze is cold, the corner of its mouth twitching almost imperceptibly. Low and slow: "You said earlier you liked this ring." Half-second pause. Then: "Is it the ring you like, or that I believe you?" On "believe you" avatar1’s eyes narrow slightly, as if waiting for @avatar2 to slip. HARD CUT. SHOT TWO (4.5-10.5s): Over-the-shoulder close-up on the @avatar2. The @avatar1’s shoulder and dark suit sit as blurred foreground on the left edge of frame. @avatar2 is centred slightly right, chest-up, the sapphire cheongsam collar and shoulder-neck line clear but not displayed. Background is blurred lamp, wood panelling and rain-streaked glass. @avatar2 does not answer at once. @avatar2 lowers its eyes to the ring first, lashes trembling faintly, @avatar2’s right hand at the bottom edge of frame brushing the rim of the tea cup, fingertips pausing. When @avatar2 lifts its eyes its lips part slightly. Softly: "A woman liking something sometimes doesn't need a reason." A very brief pause; @avatar2 looks at him, gaze unflinching: "But if you must ask, I like the way you hand it to me." Its tone is very light — like the truth, or a carefully rehearsed lie. HARD CUT. SHOT THREE (10.5-16.0s): Back to the @avatar1, closer than shot one — shoulders and face only. 85mm, background fully blurred into blocks of warm lamp and cold rain colour. An almost imperceptible slow push in. @avatar1 is silent for half a second after hearing @avatar2. @avatar1’s jaw tightens gently. @avatar1’s right hand at the bottom edge of frame slowly rotates the ring box without opening it. @avatar1’s gaze shifts from scrutiny to something deeper, as if touched by @avatar2’s words, then immediately suppresses that flicker of doubt. Even lower: "When you lie, your eyes are quieter than when you tell the truth." @avatar1 does not smile afterwards, just stares into @avatar2’s eyes. HARD CUT as the final word lands. SHOT FOUR (16.0-22.0s): @avatar2 in extreme close-up, its face filling most of the frame without distortion. Warm lamplight on one cheek, the other falling into cold rain shadow. @avatar2’s eyes widen slightly, breath shallow, lip colour restrained and vulnerable in the warm light. @avatar2 does not speak at first; its throat swallows gently once. Its inner brows lift faintly, then are quickly suppressed — as if finally pushed to the edge yet still composed. Very lightly, with a subtle tremor: "Then why do you still listen?" Half-second pause; @avatar2 glances at avatar1’s mouth, then back to his eyes: "Didn't you already know?" A faint sheen in its eyes, but no tears. HARD CUT. SHOT FIVE (22.0-25.0s): Two-person chest-up mid-close shot, no full body. They face each other across the small round table, @avatar1 on frame left, @avatar2 on frame right, the warm lamp edge and the unopened ring box between them. Frame stable, space quiet, rain pressing gently at the window. @avatar1 does not answer. @avatar1 slowly pushes the ring box toward @avatra2, stopping it at the centre of the table. @avatar2 looks at the box and does not reach for it. Their gazes collide again. Low: "Because I want to hear you say it once, yourself." @avatar2’s lips tremble faintly, as if about to admit it, then @avatar2 swallows the words. Hold their mutual gaze for 0.4 seconds. DIRECT CUT TO BLACK, no fade. SOUND: Natural spoken English dialogue, strictly distinguishing the two speakers, lip sync matched to every syllable. The man's voice low, slow and oppressively weighted but never roaring; the woman's soft and clear, gaining tension and subtle tremor in the later shots but never sobbing or shouting. Keep realistic breaths, swallows, the faint rustle of silk, the tea cup touching the table, rain at the window, and the natural reverb of a wooden interior. No background music, no voiceover, no subtitles, no exaggerated suspense sound effects.
Single unbroken handheld take throughout. @avatar is sitting pensively on a sofa inside a large bright suburban Parisian house, summer daytime, sunlight through the windows, young adults dancing and laughing around her holding drinks, loud music implied by energetic crowd movement. 0-4s: @avatar rises from the sofa empty-handed and walks toward a kitchen table lined with alcohol bottles and glasses. The handheld camera follows closely behind @avatar’s shoulder with slight natural shake. 4-8s: @avatar pours herself a glass of champagne and takes a sip. Leaves the glass on the same table after she takes a sip. The camera remains close behind @avatar. 8-13s: @avatar moves through the dancing and laughing crowd toward a large glass bay window. Some guests are standing while others sit and talk on the sofas. The camera continues following @avatar without a cut. 13-16s: @avatar steps into the garden, revealing a large pool filled with young adults swimming and laughing. The camera smoothly rotates from behind her to face @avatar. @avatar looks toward the pool, smiles. 16-20s: @avatar turns around toward the camera, removes their shirt and places it on the back of the lounge chair, then removes shorts and places them on the chair, revealing a swimwear underneath. The camera holds a medium frontal shot with slight handheld movement. 20-24s: @avatar turns around and walks toward the pool as the camera moves behind @avatar again. @avatar then breaks into a run, with the handheld camera following closely and matching @avatar’s pace. 24-27s: @avatar dives into the pool while shouting "take care," landing among the other swimmers. The camera dives into the water with @avatar and remains briefly submerged. 27-30s: @avatar resurfaces laughing and joins the people nearby. The camera rises into the air, revealing a wide aerial view of the house, garden, and pool. Bright natural summer daylight, warm sunlit color palette, candid documentary energy, and slight handheld lens wobble throughout. Background party guests and swimmers should remain generic and unnamed. No text overlays, flickering, ghosting, morphing artifacts, or hard cuts. Natural skin tones and stable continuous handheld motion throughout.
Snap, Melt, SurrenderRecreate workflow