The Habsburg AI Portrait Studies series utilizes a bespoke diffusion model retrained on its own outputs, highlighting the impact of synthetic data within AI model training. This recursive method, aimed at enhancing the model’s aesthetic qualities, was developed through a research-through-design approach that addresses the disparity in model parameter growth and training data availability. The results showcase how MidjourneyV5's average aesthetics manifest in portraits, while also serving as commentary on the implications of generative art models framed by historical inbreeding themes.
The project, Habsburg AI Portrait Studies, explores the amplification of aesthetic characteristics in generative portraits, revealing insights into synthetic image generation processes and their implications.
By deliberately retraining the diffusion model on its own output, the series reveals how such autophagous training can lead to an exaggerated representation of underlying aesthetic norms.
Aesthetic Distillation is designed to amplify the fundamental aesthetics of a model like MidjourneyV5, allowing for formalist analysis of synthetic image generation.
The Habsburg AI Portraits series serves as an artistic commentary on the limitations of training data in the context of rapidly growing model parameters.
Collection
[
|
...
]