On Machines, Users, and the Politics of Visual Style in AI-generated Images

The study of style is often overlooked, partly because it is seen as intangible, ephemeral, or ineffable. Yet, style is fundamental to understanding how visual generative AI platforms redefine the ways in which images are created, edited, used, described, and interpreted. To contextualise the continuities, novelties, and challenges of understanding visual style in AI-generated images, it is important to look at how style has been theorised across different disciplines, which the first part of this chapter does. We approach the visual style of AI-generated images as a pattern of sociocultural meaningmaking that emerges at the intersection of machines, humans, socioeconomic contexts, and visual culture. In the second part of this chapter, we then examine this intersection at three levels, developed across three sections. First, we explore how visual generative AI technologies provide the technical infrastructure to reproduce styles due to their
training on extensive swaths of data and the recent advancements in machine vision aesthetics. Second, we discuss how users can adjust and modify the default styles to varying degrees, depending on the affordances of the platform and their prompting practices. Third, we argue that photorealism has emerged as a dominant AI style, rooted in the probabilistic reproduction and
commodification of existing visual forms.