Seizing the Means of Production: Exploring the Landscape of Crafting, Adapting and Navigating Generative AI Models in the Visual Arts

Users of these models can produce diverse and high-quality visuals through meticulously written text prompts. These models mark a significant shift from the era when artists used personally-trainable generative models like Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs). LTGMs have made the generation of visuals accessible to everyone, but this shift in attention has overshadowed the practice of "model crafting", whereas artists personalize their work by experimenting with training sets, model architectures, and hyperparameters in addition to combining, adapting and manipulating pre-trained models [14]. Model crafting offered artists a sense of craftsmanship and ownership over the creative process and its outcomes.