Generative Models are powerful tools for artists and designers. With the emergence of diffusion models, prompting has become the de facto standard of interacting with such tools. However, when it comes to discovering new artistic realms, human language limits a users' capability to explore the unknown parts of these models. Research on novel ways of exploring and interacting with generative models present interesting opportunities to enhance the interactivity, especially for non-technical users. Providing users with an interactive and comprehensive map of a generative model's latent space promises an intuitive approach to exploring both, uncharted areas of a model, and unknown artistic expression.
Moritz von Schwind (Thu,) studied this question.