Abstract
Behind the popular commercial interfaces of AI image generation lies a vast network of open-source community-driven modification, where AI models are fine-tuned and shared. In this paper, we argue that studies of computer game modification (modding) offer a highly relevant precedent for analysing the structure and aesthetics of AI image generation. We have previously argued (Nelson 2023) that game software, usage agreements and patterns of community production in modding offer a rich subject for contemporary aesthetic analysis. Using existing studies on the motivations, technical affordances and community dynamics of modding (Küchlich 2005, Newman 2008, Sotamaa 2010), combined with Azuma’s formulation of ‘database animals’ (2009) and Ngai’s contemporary aesthetic categories (2012), we showed how distinct aesthetic features can be identified within the large-scale creative production of modding. This paper on AI image generation starts with two initial observations. First, the dynamics of AI fine-tuning are highly analogous to those of modding. First, the iterative fine-tuning and sharing of AI models is essentially a form of modding. Second, the aesthetic predilections of popular machine learning forms such as civit.ai strongly resemble those described by Azuma. From here, we will combine approaches from modding, digital ethnography and visual studies to first build the argument that AI fine-tuning is a form of modding, then we will construct a theory of practice and aesthetics for what is rapidly becoming a highly significant mechanism driving contemporary computer graphics and contemporary visual culture.
Original language | English |
---|---|
Publication status | Published - 7 Apr 2025 |
Event | Current Perspectives in Game Studies: Spring/Summer 2025 - City University of Hong Kong, Hong Kong, Hong Kong Duration: 7 Apr 2025 → 8 Apr 2025 https://www.scm.cityu.edu.hk/events/current-perspectives-game-studies-springsummer-2025#nelson |
Seminar
Seminar | Current Perspectives in Game Studies |
---|---|
Country/Territory | Hong Kong |
City | Hong Kong |
Period | 7/04/25 → 8/04/25 |
Internet address |