Generative Images Turn Location Scouts Into Campaign Concept Canvases
OpenAI’s marketing team argues that ChatGPT Images can move campaign exploration closer to the real-world site before a creative direction is chosen. Using a scouting photo of a Chicago billboard, the team first converted the angled facade into a straight-on design canvas, then generated competing concepts—from a brick wall opening onto a pastoral landscape to an L train emerging from the building—to assess them in context. The workflow is presented as a way to develop and compare more location-specific ideas before committing to a final composite.
Campaign exploration can begin with the physical site
The campaign premise was to show what an image model can visualize by transforming familiar billboard placements into scenes that feel less like conventional advertising and more like “surrealistic visions.” The proposed medium was deliberately large-format: billboards across U.S. cities, treated not as fixed surfaces for a finished message but as ordinary urban sites to be reimagined.
A presentation slide shows that ambition in three vertical executions installed in city settings, each labeled “Reimagined with ChatGPT Images.” The treatments are stylized and improbable rather than product-demonstrative in a literal sense. The point is not simply to put an image on a billboard; it is to make the billboard appear to alter the surrounding built environment.
That framing follows from the stated capabilities of the model: it can render text, work across languages, and turn a mental image into a visual. The marketing problem, as presented, was how to make that breadth tangible in an ad campaign. Recasting physical ad spaces gave the team a concrete constraint through which to test visual ideas.
A scouting photo became a usable design canvas
The process began not with a blank generative prompt but with a location scout. A street-level photograph showed a brick building with a large, blank white billboard mounted on its side. It was shot at an angle, which made it useful as a record of the location but less useful as a direct canvas for developing billboard artwork.
The team used ChatGPT Images to change the camera view while retaining the building’s proportions. The on-screen prompt asked for “a straight-on view of the building and whole white billboard space,” with all proportions maintained and the billboard rendered rectilinear. It also specified a close crop, a 3:4 ratio, 1920×2560 resolution, and high quality.
The resulting image, displayed as “Chicago Billboard - Straight On 1920x2560.png,” presents the facade head-on, with the billboard converted into a rectangular working area. That preliminary transformation matters because it separates two linked problems that are often otherwise handled together: first, making a real location legible as a design surface; second, deciding what creative concept belongs on it.
Rather than treating the original scouting image as merely reference material for a later production process, the workflow turned it into an editable basis for concept generation.
The model was used to compare directions, not just execute one
Once the billboard had been made rectilinear, the team could use it as a shared surface for brainstorming. The presenter describes this as filling a whiteboard with possible answers to a basic question: what should occupy that billboard?
One direction replaced the blank panel with brick matching the surrounding wall, then broke that surface apart. The prompt specified bricks “falling away into a giant portal” that revealed rolling grass hills, blue skies, and wildflowers, “as if a grand illusion.” It requested four high-quality variations at 1920×2560.
The displayed outputs show four related versions of that idea: an oval or arched opening set into the brick facade, with a bright pastoral landscape visible beyond it. The variations preserve the structural conceit—a city wall becoming a portal—while changing the composition and treatment enough to let the team judge alternatives.
A second direction moved from pastoral fantasy to a distinctly Chicago image: an L train emerging from a tunnel suspended in the building’s wall. The presenter calls it “classic Chicago,” emphasizing the visual contradiction that gives the concept force: a train appears to emerge from the building, in midair.
The two directions show the role of iteration in this process. The first uses the billboard as an illusionistic rupture in the wall; the second uses it as the mouth of an impossible piece of city infrastructure. Both begin from the same location and composition, but they test different associations, moods, and visual narratives.
More concepts can be examined before a direction is chosen
The final comparison returns to the original scouting photo and pairs it with a site composite. In the composite, the blank white billboard has been replaced by the L-train concept, with the train emerging from a dark circular tunnel in the brick facade. The image retains the street-level perspective of the actual location, making the generated work easier to assess in context rather than as a standalone illustration.
The practical claim is about the breadth of exploration available before committing to an execution. Under the earlier constraint described by the presenter, a team might identify its best three or four ideas and have time to develop only those. The new workflow is presented as an alternative to that narrowing: generate and examine more of the whiteboard, then determine what works, what does not, and where to iterate.
That does not eliminate selection. The team still needs to choose a direction and produce a final composite. But it changes when judgment enters the process. Instead of filtering ideas primarily because they are costly or slow to visualize, the team can inspect concepts in a location-specific form and make decisions from a larger set of developed possibilities.