Deep understanding of the customer journey followed with the provision of dynamic personalization has been considered to be one of the most effective digital marketing strategies. In this demonstration, based on text-to-image diffusion models, we present DiffAds, an interactive platform for personalized visual advertisement generation. To simulate the process of personalized advertising in digital marketing channels, DiffAds consists of three main stages including style preference discovery, personalized advertising image generation, and user feedback collection. The first stage presents representative images of different visual styles for the user to choose from and the second stage creates the advertising images of a commercial product tailored to user preference. Finally, the third stage allows the user to explicitly rank the generated images as the preference feedback for updating the generative models. The advertising technique behind DiffAds has been practiced in real advertising campaigns of FamilyMart Taiwan (Taiwan's major convenience store chain).