AI-generated imagery for Bonnier News

In collaboration with Bonnier News, we explored the use of generative AI for replacing stock photos in ad production.

Background
Commercial operations of Bonnier News faced challenges with stock imagery, including time-consuming searches and a lack of Scandinavian authenticity in aesthetics. The project’s goal was to explore whether AI-generated images could provide a more efficient and culturally accurate alternative.

Solution
We evaluated various text-to-image models, such as DALL·E, Stable Diffusion and Midjourney, against criteria like text-to-image alignment, photorealism and Scandinavian ambiance. The project included defining specific use cases for generated images, crafting tailored prompts for image generation, and conducting extensive human and automated evaluations. The final deliverable was a web application, which integrated two image generation models and a language model to facilitate prompting text-to-image models for creative people unfamiliar with the concept of prompt engineering.

Result
Together with the Bonnier New team, we identified advertising scenarios where AI-generated images can replace stock images. A web application was rolled out to graphic designers to assist in their creative process of ad development. An online A/B test revealed a notable preference for AI-generated images. Specifically, ads with AI-generated images showed a markedly higher click-through rate.