OpenAI marketing team generates multiple billboard concepts with ChatGPT Images

In a video posted on YouTube, OpenAI's marketing team showed how they used ChatGPT's image generation feature in ad production. The project involved large billboards in several US cities, transforming ordinary ad panels into surreal imagery. The starting point was a single scouting photo taken by a field team, but the problem was that the photo had been shot from an angle on the street, so the billboard face wasn't fully visible. The team member asked ChatGPT to render the building as if viewed straight on, so the full billboard space would be visible, and got a front-facing image back right away.

Turning Whiteboard Ideas Into Images One by One

OpenAI has previously showcased how its sales team uses ChatGPT Work in a video titled ChatGPT Work: From Meeting Briefs to Strategy Decks, Internal Demo Revealed. This new video can be seen as extending that thread of ChatGPT Work use cases into ad creative work.

Once they had the front-facing view, the team started brainstorming what to fill the empty billboard with. One direction replaced the empty billboard with a brick wall that matched the building, then had that brick crumble away to reveal a portal into another world. They pulled four high-resolution variations of this idea to compare. Another direction imagined Chicago's iconic L-train design extending out of the building into open air. For both concepts, the video showed the scouting photo side by side with the final composite, letting viewers see how each idea would actually look on site.

What Changed From the Old Process

Category	Old approach	After using ChatGPT images
Number of concepts explored	Limited to 3–4 best candidates	Every idea written on the whiteboard gets turned into an image
Decision-making	Choosing directly among a few candidates	Comparing multiple directions as actual images, then iterating
Time constraints	Months of work still only allowed exploring a few options	More directions can be explored individually in the same amount of time

In the video, the team member said, "A few months ago, the best we could do was figure out which three or four candidates were the strongest." Now, they can actually dig into every idea on the whiteboard, see what works and what doesn't, and refine it through iteration.

How the Mockups Were Made

Based on the video, the production process breaks down like this:

1. Prepare a scouting photo of the actual location where the billboard will go up.
2. If the photo is off-angle, ask ChatGPT to render it as a straight-on view so the full billboard face is visible.
3. Once you have the front-facing view, freely brainstorm concepts within the team for what to put there.
4. Turn each concept into a prompt and request several high-resolution variations to compare.
5. Pick the direction you like and keep refining details through follow-up requests.
6. Composite the finished image onto the actual site photo to check the final look.

The video links out to a full webinar recording, positioned as part of an ongoing ChatGPT Work series OpenAI is running for marketing teams.

Editor's Note

What this video really demonstrates isn't the tool itself, but how much wider the exploration can go. For ad agencies and in-house marketing teams, the bottleneck in mockup creation has rarely been a shortage of ideas — it's always been the time needed to turn an idea into something visible. You could write ten ideas on a whiteboard, but budget and time usually only allowed three or four to actually get visualized. What OpenAI's example shows is that this bottleneck has now shifted onto the image generation model itself.

The contrast with older ad production workflows makes this tangible. In the past, concept sketches or mockups went out to freelance designers, and you'd wait days just to see one direction take shape. Now that same check takes a single prompt and a few minutes, and you can move straight on to the next direction. The final polish still needs a human hand, but the speed of narrowing down early directions has clearly changed.

There's a practical takeaway here for marketing teams or smaller ad agencies too. If you generate candidate directions as images before a concept review meeting, you can spend meeting time comparing finished mockups rather than arguing over ideas in the abstract. This is especially useful for media like outdoor advertising, where you really need to see something composited onto an actual site photo to get a feel for it — the approach of correcting a scouting photo into a straight-on view applies directly. That said, the need for separate review when placing copyrighted logos or brand assets into generated images still stands.