GEN 1 · Billboard Handoff — 30s · T2Free · copies cast, sets and prompts to your workspace · you pay only when you generate
About this template
The illusion comes from a single locked wide shot from street level, where neither the camera nor the billboard content ever cuts. The video node runs Seedance 2.5 at 1080p, 16:9, 30s, and draws on three reference image nodes: an empty night street plate, a character sheet and a product photo. The prompt sets a strict scale: the corner screen is roughly ten storeys high, its LED stage is about three times the character's height, and tiny pedestrians and taxis stay in the lower third as a size cue. The camera is written as jittery handheld footage from someone in the crowd, with no zoom, push-in or reframing, so when the character leans out past the building edge with the jar, the forced perspective reads as real.
Consistency is anchored by a dedicated character model sheet generated with GPT Image 2.5 Flare at 2K, 4:3. The sheet shows Elena from several views on a cream background, with a title block, a short profile and labelled colour swatches for hair, eyes, skin and outfit. It also covers big expressive faces such as a wide laugh and a raised-eyebrow look, which match her playful performance on the billboard. Because the video node references this sheet rather than describing her from scratch, she looks the same while getting dressed, dancing, turning and handing over the product.
The canvas holds three image nodes and one video node. The street plate keeps the LED stage empty so any character can be placed in it, the Elena sheet defines the performer, and the product node shows a glossy black jar with slim gold lettering, angled so the label is only partly readable. To remake it, replace the jar image with your own product, regenerate the character sheet for your model, and keep the scale and camera instructions in the video prompt unchanged. Then generate, trim the result on the timeline and export to mp4 or FCPXML.