Anyone who lives or works in a building meets the same screen several times a day. In the two or three minutes spent waiting for the lift, a spot plays over and over, fifteen seconds at a time, with no fast-forward and no channel change. Attention does not necessarily stay on the screen; many people look back at their phones within seconds. That is both the opportunity and the problem: exposure is frequent, each glance is brief, and familiarity can turn into irritation when one cut runs for too long.
Fifteen seconds holds less than people expect
Common placements use short formats such as 15 or 30 seconds and repeat them according to the media schedule. There is no room for a build-up, so the brand and the point have to be readable at the first glance. Worse, the spot runs many times a day. What feels fresh on the first pass becomes background noise by the tenth. And since screens are sold by building and by district, office workers and parents in a residential compound should not be watching the same cut.
The hard shots are the ones AI handles well
To make someone look up, the image has to carry weight: a scene that could not be filmed, a product with visible mechanical detail, a cleanliness that live footage rarely reaches. Those shots are expensive in a traditional pipeline, and most of the money is spent before the camera rolls, with no guarantee the result will be usable. Generative video turns that upfront cost into a per-clip one. You write a prompt, render a few options, and change what does not work. Seedance 2.5 arrived in July 2026, generates a full 30 seconds in one pass, and accepts up to 50 reference images, videos and audio clips combined; what a given account can use depends on the current documentation.
The style no longer follows the budget
Elevator screens carry very different kinds of content. In a residential compound a real person talking to camera is the easiest to accept; 2D animation suits products for mothers and children; cars and home appliances need 3D to explain how they work. In the past the choice followed whatever you had: live action meant scheduling people and places, animation meant finding another team. Now the same selling point can be tried as a digital presenter or a UGC-style shot, then as 2D or 3D, and the version that lands better is the one that ships.
When one spot gets cheap, five become possible
The fresher the material, the less the audience minds; the faster it is refreshed, the higher the bill. Those two forces used to pull against each other. Leave one spot on air for two weeks and the first week sells while the second only spends patience. Making more versions used to mean shooting more. With generative production the second and third version cost far less, because characters, sets and brand assets get reused; a new opening, a different order of selling points, even a different style is just another run. Media buying can then rotate material on a schedule instead of pushing one film to the end.
Cheap still has conditions
Logos, product appearance and packaging copy have to match across versions, because a small mistake gets shown again and again. If real people, voices or other people's work appear, clear the licence first; commercial placement is less forgiving than a personal post. Someone still has to pick the clips, grade them, check subtitles and audio, and property managers decide what may be played at all, so format and aspect ratio are best settled before upload.
Most elevator advertising problems come back to the material: whether it earns a glance, and whether it still works on the twentieth viewing. Once production cost and turnaround drop, creative range and versioning become practical. To see how different teams handle this, or to work out what your own brief should say, start with the AI advertising production service and look at the work already there.
