A two-step recipe for the talking-product format: an image step turns the uploaded packaging into a character, and a Seedance 2.5 step gives it a 25-second monologue in one continuous take with a moving camera. WHAT TO EDIT WHEN YOU RUN IT: the monologue in step 2. The words between the { } braces are an example — replace them with your own and keep the braces. Budget roughly 65-70 words for 25 seconds; more than that and the model compresses the closing lines until they are swallowed, which is the single commonest way this format fails. WHY THE SPEECH CLOCK IS IN THE PROMPT: without it the model runs the monologue too fast and leaves four or five seconds of dead air at the end, or crushes the last two sentences into two seconds. With the clock written out the pauses land within a few tenths of a second of where they are asked for. WHY THE HANDS ARE CONSTRAINED: on an early render of this format the direction said 'raises a single index finger' and the model raised the middle finger instead, and held it for ten seconds. A stubby cartoon hand cannot isolate one digit reliably, so both steps forbid pointing and finger-counting and use whole open palms. KNOWN WEAKNESS — READ BEFORE USING A REAL BRAND: step 1 does not always respect the printing on the packaging. On the example shown here it rotated the label ninety degrees and added two words that were not on the input photo. The prompt now forbids this explicitly, but check the step 1 output before spending on step 2, and re-run step 1 if the wording or its orientation has drifted. THE PROP IN STEP 2 IS OPTIONAL: the fizzing glass suits an effervescent tablet and nothing else. It is left in the prompt as a worked example of the technique, because a prop that visibly changes from the first frame to the last is a clock the viewer can read and it proves the take was never cut. Swap it for something that changes in your own product's world, or delete the block. PREVIEW: produced by this chain at 720p. Step 2 runs at 1080p for a finished ad.
Drag the clip sideways
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.
A 22-second vertical 9:16 talking-product comedy ad. Photorealistic 3D character animation, cinematic. ASSET MAP — every reference is bound here and cited by tag below: @image1 — the magnesium product. This is the product's truth: its shape, material, colour, closure and every word printed on it. @image2 — the same product wearing its cartoon face and arms. This is the character design. PRODUCT LOCK — HIGHEST PRIORITY: the container is exactly the one in @image1. Preserve its shape and proportions, its material and colour, its cap or closure, and its label artwork exactly. Every word, number and symbol printed on it stays as photographed — same spelling, same typeface, same size, same position. Never add, remove, translate or re-typeset any lettering, brand name, logo, badge or seal. Wherever this text and the attached images disagree, follow the images. CHARACTER LOCK: the face and arms are exactly the ones in @image2 — same eyebrow weight and shape, same eye style, same mouth, same arm colour and thickness, same placement on the container. Do not redesign them. THE EYEBROWS ARE THE PERFORMANCE — two thick, solid, matte-black eyebrows shaped like bold brush strokes, clearly separated from the eyes and sitting well above them. They are the single most expressive element in the entire shot: they tilt, arch, furrow, raise and flatten on every emotional beat. They are never thin, never a mere eyelid line, and never absent from any frame. HANDS STAY CLEAR OF THE LABEL: the arms gesture beside the container and out to its sides, but never cross in front of its printed text. Every word on the product stays readable in every frame. THE BODY NEVER DEFORMS: the container does not stretch, squash, bend or hop. It stays standing on the counter. Only the face and the arms move. SET: a dark kitchen at three in the morning. The container stands on a stone counter. The only light is a cool under-cabinet strip and a faint glow from a doorway behind. A blurred kettle, a mug and a plant deep in the out-of-focus background. Quiet, still, nobody else in the room. CAMERA: one continuous shot, no cuts. It opens WIDE — the whole dark kitchen, the container small and alone on the counter — and pushes in slowly and steadily across the entire clip, arriving on a medium close-up of its face by the final line. The move is smooth and never stops. No handheld shake, no pan, no whip. VOICE: one voice, dry, weary and completely deadpan — something that has been waiting all day to say this. Not cartoonish, not sinister, not whispered. Ordinary speaking pace, with real pauses between sentences. PERFORMANCE AND LINES — it speaks these words, in this order, word for word, exactly once: 0–4s: wide shot. Arms crossed, eyebrows flat and low, already unimpressed before the clip starts. "Three in the morning. Calf cramp. And suddenly, you remember I exist." 4–9s: uncrosses its arms, one hand turning up in a small matter-of-fact shrug. Eyebrows level, plainly stating a fact. "I'm magnesium. I'm involved in hundreds of reactions in your body. Quietly. Every day." 9–14s: leaning into the grievance, eyebrows angling sharply down into a furrow, one hand pointing straight at camera on each question. Midway through this stretch a single tiny white spark flares off the cap for half a second and dies — it does not react to it. "That eye twitch last Tuesday? Restless legs at night? Your body was asking." 14–20s: one hand up flat in a stop gesture out to the side, then both hands opening outward. The eyebrows travel from firm and level to raised and almost pleading. "I'm not a painkiller. Don't wait for the cramp — take care of your magnesium first." 20–22s: the arms drop to its sides, the eyebrows settle flat, the eyes glance away, and it holds there as the camera settles on its face. "...I'll be here." AUDIO: quiet night room tone throughout, a faint fridge hum. One soft electrical tick on the spark. The voice. No music bed, no score under the dialogue. NEGATIVE: on-screen text, captions, subtitles, a watermark, a brand name, an invented logo, added or altered lettering, garbled label text, a hand covering the label, thin eyebrows, missing eyebrows, a human, hands belonging to a person, a second container, a second character, a face on any other object, photorealistic human eyes, teeth, lips, stretching, squashing, bending, hopping, the container leaving the counter, camera cuts, handheld shake, flickering light, horror atmosphere, sinister tone, spilled capsules.
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.
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.
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.
Break the IceRecreate workflow