[en] In this short talk, we explain how the text-to-image model DALL•E works, and how it perceives a given theme, namely Susanna and the Elders, to generate new images. We examine the general culture of the machine, and how it can perform compositional variations.
