One Product Photo, One 15-Second Ad: Inside Pollo AI's Workflow

On October 8 local time, OpenAI published a case study on Pollo AI, a service that plugs OpenAI models into its own product to help users create content. Feed it an idea and a product photo, and it produces image and video ads. The case study walks through the full arc: sketching out an ad’s storyline and scenes, then producing posters and short video spots.

According to Pollo AI, the service has more than 26 million users. Content creators make up the bulk of them, and use by marketers and business owners is growing as well. The work shown off includes posters, storyboards that lay out shots before filming, and campaign imagery. The case study focuses on producing images and video to promote a product.

How an Idea Becomes an Ad

Pollo Agent takes a rough idea and some visual reference material, then generates a storyline, a scene breakdown, and a script. It is designed to be usable well before every shot has been nailed down. Write out what you have in your head, or drop in a reference image, and the agent assembles the narrative and what belongs on screen.

Different models handle different parts of that process. GPT-5.6 classifies incoming tasks. GPT-6 Astra builds out complex narratives and plans scene revisions. Images come from GPT-Image-2.5. In other words, one model routes the work, another fleshes out the story, and a third draws.

A 15-Second Video From a Single Product Photo

Photo to Video Ads, part of Marketing Studio, turns one photo plus a prompt into a short video ad. The feature targets users who want to start from a product shot and end up with a video spot. The perfume ad featured in the case study runs roughly 15 seconds.

Starting from a prebuilt video format is another common path. Pollo AI says 30% of creation sessions begin with a video template or a trending format.

How Much Less Time Do Users Spend Choosing Models?

Pollo AI says the time users spend picking or switching models has dropped by more than 50%. The company says it evaluates models on output quality and consistency, the range of styles they can express, and speed and cost.

The through line here is a pipeline that takes a product photo and a simple idea, runs them through ad planning, and delivers images and video. That is useful not only for content creators but also for marketers and business owners promoting their own products. The real test in production use is whether it cuts creation time and cost while holding the line on the quality and style you want.