Generate an action commercial video for noodle products.

2 steps2396.7K

Ultra-realistic premium beverage commercial, 20 seconds, 3:4, continuous cinematic sequence built from 8 shots, @product as the central hero throughout. Reference @product for label, bottle shape, cap color and liquid color — keep identical across every shot. Style: crystal-clear liquid, realistic condensation physics, natural premium lighting, clean packaging reflections, luxury commercial photography, shallow depth of field, subtle motion blur on fast liquid/ice movement, 8K detail. 0–3s: chilled @product stands on wet natural stone, condensation beading on the surface; slow push-in. 3–5s: extreme close-up on the cap/seal opening — crisp mechanical motion, a small hiss of released pressure. 5–8s: liquid pours from @product into a transparent glass, smooth continuous stream, no splashing outside the glass. 8–10s: swirling liquid inside the glass, tiny bubbles rising, macro detail on the surface tension. 10–13s: Garnish pieces drop beside @product in slow motion, natural physics, soft impact. 13–15s: ice cubes fall into the drink, controlled splash with droplets catching the light, no liquid leaving the glass. 15–18s: extreme macro on condensation droplets running down @product's surface, natural gravity and pooling. 18–20s: final hero shot — @product and glass centered, surrounded by garnish and a frozen mid-air water splash; camera holds still for the last second. Audio: tropical water ambience, refreshing ASMR-style liquid/ice sounds, soft premium beverage-ad score building toward the final hero shot. Negative: no visible human hands/faces unless added intentionally, no liquid color/label inconsistency between shots, no unrealistic splash physics, no generic stock-footage look, no subtitles.

2 steps66544K

Live-action rustic product film with glowing golden line-drawings rising out of the steam above the mug.

2 steps1999.5K

Generate a video of your avatar with 3 different product.

6 steps33718.1K

Two steps, one generation. Step 1 is the visitor's photo (@image1): shot at eye level, vertical or croppable to 9:16, with 3 or 4 clear objects in view (vehicles, boats, furniture, anything) and at most one small person, no dense crowd. Streets work best because everything already moves. Step 2 is one unbroken 15s 9:16 Seedance 2.5 render with @image1 attached as the first frame. The prompt never names a city, a brand or a specific object, it picks the largest object, a smaller moving one, a third one and a person if present, so the same recipe works on any photo. Known traps: dense crowds make Seedance 2.5 fail, crop them out before uploading; a face filling the frame gets blocked by moderation, keep people small; the original viral version shouts Biu! on every hit, this recipe swaps that for a sparkle reveal chime so nothing has to be translated.

2 steps43620.5K

Seoul street at night, medium close-up. @avatar walks away from the camera along a narrow street lit by defocused neon signage and shop light, wet asphalt throwing reflections. The camera follows from just behind at @avatar's eye level, handheld with subtle natural instability — never jittery. @avatar turns its head back over its shoulder to glance into the lens and smiles, lifts its right hand and brushes its hair back from her face, then turns forward again and continues walking, drifting to the left of frame. Each of these happens exactly once, in that order — no repeated glances, no looped gestures.

2 steps19326.9K