PRODUCT LOCK — CRITICAL: @image1 is the exact bottle. Photocopy its shape, cap, body colour, wordmark and every label line exactly as printed. Do not redesign. Do not turn it into another brand. One bottle only. Ignore any background or surroundings in @image1 — the film is a clean white studio. 10 seconds, 9:16, luxury beauty commercial. No people, no hands covering the bottle, no on-screen captions. AUDIO: soft water, liquid, lather, hair. No speech. No lyrics. 0.0–1.5s WATER REVEAL. Extreme macro, crystal-clear droplets falling slow-motion on soft white. Sparkle, forward push, shallow DOF. 1.5–3.0s HAIR. Close-up long silky hair flowing slow-motion, realistic strands, studio highlights. No face. 3.0–4.5s SHAMPOO TEXTURE. Creamy white shampoo flowing from THIS bottle (@image1), glossy ribbon, real viscosity. Bottle undistorted, label readable. 4.5–6.0s LATHER. Cream mixes with water into soft foam, tiny bubbles, droplets, macro. 6.0–7.5s SILKY HAIR. Camera glides through shiny hair swaying. No face. 7.5–10.0s HERO. The exact bottle from @image1 standing upright on glossy white. A few droplets, slow push-in. Last second: hold, sharp, centered, undistorted. NEGATIVE: another brand's bottle, distorted bottle, altered label, extra bottles, people, hands, watermark, subtitles, clutter, floating bottle.
Drag the clip sideways
Your avatar blooms into a lush red-flowering garden bush — head becoming one big red flower, aware and reacting, until no skin is left and the figure reads as a real leafy shrub of petals and green foliage blending into the hedges. Music only, loud. See the video step for the full prompt.
Smooth natural motion, clean commercial fashion-ad aesthetic, crisp detail, luxury catalogue presentation, soft studio lighting with realistic shadows under the figure and under the pieces. @image1 is the layout and the first frame, and it is an absolute lock. The entire layout stays exactly as shown: same background, same item panel, same captions, same crown icons, same wordmark, same framing, same position and same scale of every element. Face, hair, body, proportions, garments, materials and colours stay exactly as they are in that frame, with no drift or reinterpretation in any frame. Camera locked off on a tripod. No pan, tilt, zoom, tracking, dolly, handheld movement or reframing. The figure turns one full circle on the spot the way a fashion model turns on set, and the pieces inside the panel turn with them. Beats, each given equal weight: They begin in the pose of the first frame, then start to turn, stepping their feet around naturally and shifting their weight from one leg to the other, arms moving freely and shoulders relaxed. They continue round through profile, the turn unhurried and even. They come around to the back and glance over their shoulder toward the camera. They continue round through the other profile, still stepping the turn out on the spot. They return to face the camera and settle into a new, different confident pose, and hold it. Across the same turn, every piece inside the item panel rotates in place on its own vertical axis like a luxury product display render, revealing front, side and back. Each piece stays centred in its own slot, keeps its size, and completes exactly one full rotation, ending in its original orientation. The figure stays in the same spot and at the same size in frame throughout, feet staying on the floor with the contact shadow following them. Every garment continues around the side and the back, complete and unchanged, with each trim, fastening and fold consistent all the way round. Natural posture and realistic cloth physics: the garments, their hems, cuffs and edges, anything worn at the neck or the ankle, and the hair all swing and settle with the motion, then come to rest. The panel, its divider lines, the captions, the crown icons and the wordmark stay perfectly fixed and stay sharp and legible for the whole clip. Only the figure moves and the pieces rotate. No camera movement, no cuts, no identity drift, no face change, no garment substitution, no change to the panel layout or its type, no drifting or resizing of the panel, no sliding feet, no warped hands, no extra limbs, no disappearing accessories, no added scenery, no extra text, no logos, no watermark.
integrated_multimodal_description: A photoreal fashion spec sheet that animates, held in one locked-off vertical frame. Warm off-white paper ground, soft even warm-neutral light, no cast shadows, everything in focus, no colour grading, no film grain. The frame never pans, tilts, pushes or zooms; every change happens inside a still frame. REFERENCES — the attached images outrank every word here. @image1 is an absolute character lock: face, body, hair and everything worn in it are taken from it exactly, with no drift or reinterpretation in any shot. @image2, @image3, @image4 and @image5 are the four wardrobe lineups; reproduce each one exactly as composed — the three views, their order and spacing, the stance, the annotation rules and labels, the bottom-left wording and its colour, and the palette chips. Nothing is redesigned, recoloured or re-laid-out. This prompt names garments only by slot and never describes them; the lineups are the only source of garment design, colour and material. Shot 1: The figure from @image1 stands full length, front on, centred, alone in the frame. They turn slowly and evenly on the spot, front to profile, feet pivoting under them. Small registration ticks blink on and off in the empty margins around them. Shot 2: The turn continues through the back view and returns to front on. As they settle, the measurement overlay draws itself on from the top of the frame downward: rules extend, end ticks snap in, leader lines run out to the margins, and the numeric callouts fade up in place. The unit caption settles last. Shot 3: Cross-dissolve. The lineup in @image2 resolves in place of the single figure, all three views arriving together. The side rules draw down the edges, the dimension brackets snap across the top, and the bottom-left wording and the palette chips fade up. Shot 4: Cross-dissolve to the lineup in @image3, the three views ghosting in over the outgoing ones before they clear. The same annotation build runs: side rules, then brackets, then the wording and chips, the wording arriving in its own colour. Shot 5: Cross-dissolve to the lineup in @image4. The same annotation build runs in the same order, the wording again in its own colour, the chips landing one after another from left to right. Shot 6: Cross-dissolve to the lineup in @image5. The same annotation build runs a final time, and the frame holds steady once the last chip has landed. Every movement runs forward in normal time: the subject always travels in the direction their body faces and nothing moves in reverse. Between the three views inside any one lineup, nothing about the person changes. Do not generate any text other than what the attached lineups already carry. No logo, no watermark, no subtitles, no captions, no brand names.
Upload an outfit photo and an avatar. Step 2 generates a clean outfit-only image; step 3 magically reveals your avatar wearing that exact outfit, keeping the same face. Music only, loud.
Your avatar turns into fully transparent, see-through water and then suddenly collapses and splashes down into a clear puddle on a sunny poolside. Music only, loud. See the video step for the full prompt.
SHOOT 1 ELARA LOCK: @image1 is Elara — photocopy her exact face, hair and eyes exactly as in @image1, gold hoops, gold chain. Same person every frame. Never change their hair colour. Do not beautify. ROOM: the apartment in @image1, exactly as shown, at golden hour. Hallway beyond the door is the same home, warm wood, off-white walls. LOOK 1 (this whole take): exactly the outfit worn in @image1. SCENE: Elara films herself on her own front-facing camera while walking a warm dim hallway and through an open door into the sunlit living room, talking the whole time. Camera is her POV of herself, arm's length. Opens as she brings the camera up. One unbroken take. She crosses into late-afternoon sun, turns, stops in the light for the last line. 9:16 vertical. ~10 seconds. Real-time. No cuts. CAMERA: front-facing phone, swings up from below, holds ~50cm from face, soft bounce on steps, focus on face. ACTION: already walking. Delivers the line into the lens. One brief hair fix. Steps through the door into volumetric sun. Slows and stops for the last words, looking at the lens. VOICE, English, casual then quieter: "I don't know if I should say this, but the thing that scares me about 2026 is that nobody can tell I'm not real." AUDIO: NO MUSIC. Diegetic only — her voice close, footsteps on wood, fabric, room tone. No subtitles. No second person, no phone visible in frame, no logos, no UI. Living micro-motion every frame. Sun from camera-right in the room. SHOOT 2 ELARA LOCK: @image1 is Elara — photocopy face, hair and eyes exactly as in @image1, gold hoops. Same person every frame. Do not recast. ROOM: apartment in @image1, exactly as shown, late-afternoon sun from camera-right. Locked-off vertical UGC. No studio. @image3 = the sweater. Photocopy it exactly as in @image3 — knit, colour, collar, buttons and hem. @image5 = the shirt, exactly as in @image5. @image4 = the skirt, exactly as in @image4. Any label inside the back neck of the @image3 sweater stays exactly as it is; never invent a label. LOOK 1 (first half, still wearing): the outfit from @image1. LOOK 2 (after the cut): the @image5 shirt UNDER the @image3 sweater (collar and cuffs peeking) + the @image4 skirt + tan loafers. FORMAT: 9:16, 15 seconds, timed multishot, exactly ONE HARD CUT at 0:09. Instant cut. Same camera, room, light. Only wardrobe changes. She never speaks. CAMERA: fully locked off, low tripod, head-to-toe, one framing both sides. Phone autofocus hunts when the label comes close. 0:00–0:09 SHOWING. Already in motion. Holds the folded @image3 sweater, the @image5 shirt and the @image4 skirt. Lifts the sweater, lets it fall open, snap, show front and back. Drapes it, drops the skirt open so its details read, then the shirt. Steps in, folds the sweater inside back neck to the lens so its own label fills the frame. Focus hunts then locks on it. HARD CUT. 0:09–0:15 WEARING LOOK 2. Already dressed — no changing shown. The @image3 sweater over the @image5 shirt, the @image4 skirt, loafers. Tugs the hem, hand down the skirt, hair behind ear, weight on one hip, quarter turn, small closed-mouth smile. Mouth stays closed. No speech. AUDIO: NO MUSIC. Fabric rustle, loafers on parquet, room tone. No subtitles. No second person, no phone, no mirror, no extra logos. Fully clothed every frame. Living motion. SHOOT 3 ELARA LOCK: @image1 is Elara — photocopy face, hair and eyes exactly as in @image1, gold hoops. Same person every frame. Do not recast. ROOM: apartment in @image1, exactly as shown, late-afternoon sun from camera-right. Locked-off vertical. Same framing as the previous try-on. @image2 = the dress. Photocopy it exactly as in @image2 — fabric, colour, straps, neckline, detail and length. STARTING OUTFIT (first half): still wearing LOOK 2 — the @image3 sweater over the @image5 shirt, the @image4 skirt, tan loafers. AFTER THE CUT: LOOK 3 — only the @image2 dress. Tan loafers peek under the hem. FORMAT: 9:16, 15 seconds, exactly ONE HARD CUT at 0:08. Instant. Same camera and room. Only wardrobe changes. She never speaks. NO brand label close-up in this shoot — do not invent a brand label on the dress. CAMERA: locked off, low tripod, head-to-toe. 0:00–0:08 SHOWING THE DRESS. Already in LOOK 2, already lifting the @image2 dress. Lets it fall open full length, hem swaying. Small lift so its details read. Rotates to show the back, fabric swinging. No neck-label insert. HARD CUT. 0:08–0:15 WEARING THE DRESS. Already in the @image2 dress, no changing shown. Hand down the fabric, hair behind ear, hip pose, then one slow light pirouette — skirt and hair lift, not a trained dancer. Lands, closed-mouth smile. Mouth stays closed. No speech. AUDIO: NO MUSIC. Fabric swish, hem, loafers on parquet, room tone. No subtitles. No second person, no phone, no mirror, no logos. Fully clothed every frame. Living motion.
Your avatar turns to stone from the feet up, aware and struggling, ending as a weathered statue among the rocks. Music only, loud. See the video step for the full prompt.
Your avatar bursts into flames and rises as a fire-wielding villain, unharmed and menacing, wreathed in towering flames on a night city street. Music only, loud. See the video step for the full prompt.
@image1 fixes the object's identity: shape, proportions, construction, materials, hardware and colour are taken from it and must not drift, restyle or change at any point. The soiling seen in @image1 applies to the opening shot only. @image2 is the storyboard and is followed exactly — the same three shots in the same order, with the same framing, camera distance, hand positions and lighting. Style: photoreal macro cinematography, shallow depth of field with the focus plane riding the object, low-key directional light, cool neutral grade with warm specular highlights on the object's surfaces, fine natural grain, restrained handheld micro-movement throughout. Shot 1 — overhead close on the bench. Two hands in black nitrile gloves scrub the soiled object with a small detail brush through grey soapy foam, working along its length in short even strokes and lifting grime out of its seams. The camera eases in slowly from above. Shot 2 — cut to the same gloved hands holding the object up to camera, now fully cleaned. They turn it through a slow quarter rotation so the light travels across the cleaned surfaces. The camera stays where it is, focus riding the object as it turns. Shot 3 — cut to the cleaned object resting on the clean dark surface, hands out of frame. The camera pushes in slowly and settles, light raking across the object, holding to the end. Audio: close ASMR field recording — brush bristles working over the surface, wet foam shifting, small contact clicks against the bench, water and cloth, faint room tone. No music, no voice. Do not generate text, subtitles, captions, logos or watermarks. Do not substitute, redesign or restyle the object. Do not add tools, props, hands or people beyond those described. No cuts other than the two described.
Your Magnesium Talks BackRecreate workflow