Blog

The fake-OOH ad playbook: why giant CGI products win social

Why fake-OOH ads — giant CGI products dropped into real streets — stop the scroll and get shared, and the six decisions that make one hold up.

When Jacquemus sent handbags the size of delivery vans driving through the streets of Paris, and Maybelline ran a London train through a giant mascara wand, neither brand built a thing. No permits, no cranes, no set — both clips were CGI composited into real street footage, and both traveled far beyond either brand's owned audience. The format has a name — fake out-of-home, or FOOH — and a growing catalog of imitators, because underneath the spectacle the playbook is surprisingly repeatable.

We covered what FOOH is and where the trend came from in a separate explainer. This is the practitioner's half: why the giant-CGI-product formula keeps winning on social — the psychology is more specific than "it looks cool" — and the six decisions that separate a clip people share from a clip people scroll past.

Why a giant product stops the scroll

Your brain is a prediction machine. It has seen ten thousand street scenes and knows, without asking you, what belongs in one — cars, curbs, pedestrians, scaffolding. A product scaled to the size of a building violates that prediction in a scene that is otherwise perfectly ordinary, and the violation gets flagged before taste or intent get a vote. You don't decide to look at a six-storey perfume bottle unfolding against a facade. You've already looked.

Then comes the beat that makes the format special: for half a second, the clip is unresolved. The street is demonstrably real — that's clearly a real intersection, real daylight, real traffic — so the giant product might be too. Stranger installations have happened. That "wait, is that real?" moment is doubt, doubt is watch time, and watch time is distribution. A conventional ad announces itself instantly and gets scrolled on reflex; a fake-OOH clip withholds its verdict, and the viewer stays to reach one.

Note what's carrying the illusion: everything except the product. The trick only works because the rest of the frame obeys the rules — flat unglamorous light, believable shadows, pedestrians who react like bored humans rather than extras. One impossible thing, in a world that is otherwise completely obedient.

Why it gets shared — and why the brand survives the trip

The sharing mechanic is just as specific. People don't share fake-OOH clips the way they share a joke; they share them the way they share a riddle — to ask. "Is this real?" is the caption, the group-chat message, the comment war. Every repost recruits new people into the argument, and the argument is free distribution that conventional creative almost never earns.

And unlike most viral formats, the brand survives the trip. The usual tragedy of viral creative is that the funny part travels and the brand falls off in transit. In a fake-OOH clip that can't happen, because the product is the spectacle — the thing being shared is your bottle, your sneaker, your lipstick, at landmark scale. There is no version of the clip without the brand in it.

One honest caveat before the playbook: this is a creative swing, not a reliable performance format. Most attempts don't pop. It's an organic-first, top-of-funnel awareness play — not a replacement for a conversion ad with a hard CTA — and the ones that do pop are usually a strong idea more than a strong render.

The playbook: six decisions that make or break it

1. Pick a place people recognize. A landmark, or a street so typical it might as well be one. Familiarity does half the persuasion: the viewer's own memory testifies that the location is real, which extends credit to the thing you've put in it. An invented fantasy street surrenders that advantage.

2. Do exactly one impossible thing. The product rolls down the avenue, pours over the plaza, unfolds against the building, blinks its lashes. One clean action. The moment a clip stacks a second surreal beat, it stops reading as "installation I stumbled onto" and starts reading as "ad" — which is the one thing it must never read as.

3. Let everything else obey physics. Neutral daylight, shadows that land where shadows land, traffic that keeps moving, passersby who glance or film or ignore it the way real people would. The mundane sells the monster. This is also why favoring rolls, pours and reveals over throws and drops matters: an object in free flight is where AI physics still visibly breaks, so choreograph motion that stays carried and grounded.

4. Scale up a real product photo, not a description. At building scale, every proportion error in your product is a billboard-sized proportion error. Ground the render on an actual product photo so the shape, label geometry and color survive the scale-up. A text prompt describes your category; a reference image pins your product.

5. Shoot it like a passerby. Locked-off or lightly handheld, phone height, one position — bystander grammar. The clip's fictional premise is "someone filmed this on their way to work," and a swooping drone move breaks that premise instantly. If the camera could not have been held by a person standing on that corner, the found-footage frame collapses.

6. Composite the logo and captions in post. Generative models still garble wordmarks, and at FOOH scale your logo is the largest text object in the scene. Render the scene clean, then overlay the real logo and any caption on the finished footage. Crisp type on one impossible object, photographed plainly — that's the whole formula.

Launching it: organic-first, then amplify

Post it native and organic, with no hard CTA — a caption that invites the question ("spotted this morning…") rather than answering it. Let the comments argue; the argument is the reach. When people or press ask directly, say it's CGI — the reveal has become a second story beat for the famous examples, not a liability, and some platforms now expect AI-generated content to be labeled anyway. Never claim it's a real installation. Then, if the clip earns real traction, cut the paid version: same footage, an end card and CTA composited on, run as an awareness layer while the organic conversation is still warm.

Make one with the Faux OOH format

This is a named format in Hermoso's studio — Faux OOH: a giant product in a real street scene — so the playbook above is what the render pipeline already does: it grounds the shot on your real product photo so identity survives the scale-up, composites your logo and captions in post instead of letting the model paint them, and routes the job across the current top video models (Seedance 2.0, Veo 3.1, Kling 3.0) depending on the shot. Its Ad Spy can also show you whether anyone in your category is already running FOOH-style creative before you spend a credit. The free plan includes 250+ earnable free credits, no card required — a fake-OOH experiment costs an idea and a few minutes, not a VFX invoice. Try one on your own product and see if the comments start arguing.

Frequently asked

Why do giant CGI product ads work so well on social?

They break the brain's prediction of a familiar scene — a building-sized product in an ordinary street gets attention before taste gets a vote — and then hold it with doubt: for a beat, viewers can't tell if it's real. That doubt drives watch time, and the "is this real?" question drives shares and comment arguments, which are free distribution.

What makes a fake-OOH clip believable?

One impossible thing in an otherwise obedient world: a recognizable location, exactly one surreal action, physics and light that behave everywhere else, a camera a passerby could plausibly have held, a real product photo as the reference so the product stays true at scale, and the logo composited in post rather than painted by the model.

Do I need to film a real street?

A few seconds of real phone footage of a street or square is the strongest foundation, but a generated location plate can work — especially for a locked-off or slow-drift camera. Real or generated, the location should read as ordinary; the product is the only thing allowed to be impossible.

Should I admit the ad is CGI?

Yes — when asked, say so plainly. The reveal has become part of the story arc for the format's famous examples, and some platforms now expect AI-generated content to be labeled. Let the caption invite the question, never claim a real installation, and treat the confirmation as a second beat of the campaign.

How much does a fake-OOH ad cost to make?

The old way was a VFX invoice — camera tracking, 3D modeling, compositing, days of work. With AI video it's a standard render measured in credits and minutes: on Hermoso, Faux OOH is a built-in format, and the free plan includes 250+ earnable free credits with no card required, so a first experiment costs an idea rather than a budget.

Faux OOH is a built-in format in Hermoso's studio — grounded on your real product photo, logo composited in post, routed across the current top video models. The free plan has 250+ earnable free credits, no card required.

Start free   ← All posts
Mara Vivanco is Creative Research Lead at Hermoso, where she studies what makes ad formats travel and translates it into the studio's rendering pipeline.