Modulai collaborated with Bonnier News to explore whether generative AI could replace stock photos in ad production. By evaluating multiple text-to-image models and delivering a purpose-built web application, the project enabled graphic designers to generate culturally relevant imagery faster — and an A/B test showed that ads with AI-generated images achieved a markedly higher click-through rate.
Challenge
The commercial operations of Bonnier News faced challenges with stock imagery, including time-consuming searches and a lack of Scandinavian authenticity in aesthetics. The project set out to explore whether AI-generated images could provide a more efficient and culturally accurate alternative for ad production.
Solution
Modulai evaluated various text-to-image models — including DALL·E, Stable Diffusion, and Midjourney — against criteria such as text-to-image alignment, photorealism, and Scandinavian ambiance. The project included defining specific use cases for generated images, crafting tailored prompts, and conducting extensive human and automated evaluations. The final deliverable was a web application integrating two image generation models and a language model to help creative staff unfamiliar with prompt engineering produce high-quality imagery.
Tools
The team evaluated DALL·E, Stable Diffusion, and Midjourney for image generation, alongside a language model used to assist with prompt crafting. Both human and automated evaluation pipelines were developed to assess output quality. The final solution was delivered as a web application integrating the selected models into an accessible interface for graphic designers.
Value created
Together with the Bonnier News team, specific advertising scenarios were identified where AI-generated images can replace stock photos. The web application was rolled out to graphic designers, and an online A/B test revealed that ads featuring AI-generated images achieved a markedly higher click-through rate compared to those using stock photos.
Learn more
