This thesis study explores how a Style-based Generative Adversarial Network (StyleGAN) can be employed as a tool to design album cover artwork. The training dataset was created by scraping and curating around 150.000 album cover artworks from the open source music sharing community Discogs, including the accompanying metadata. These data were used to train an adaptation of NVIDIA’s StyleGAN (2019) to generate high resolution images. Questions about design agency with the use of such a tool are explored and an interface to navigate the latent space of the network is introduced. Finally, a survey has been done to review if the generated album cover artwork is considered good by designers: which aspects still leave room for improvement, and which aspects could be useful tools for exploration and decision-making in the design field.