BlinkArt: Using Change Blindness For AI-Generated Image Modifications in Artworks

Change blindness, the failure to detect substantial changes, is a known phenomenon that demonstrates one of the limits of human visual perception. Generative image models can now synthesize and edit photorealistic scenes on demand, with prompt-level control over the magnitude, semantic category and salience of a change while observers struggle to distinguish their output from real photographs. The availability of such fluid image adaptations brings fascinating possibilities when combined with change blindness, with respect to dataset sizes and a broadening of the range of testable changes. We present BlinkArt, a system which swaps AI-generated peripheral modifications into displayed images precisely during natural blinks. The system can autonomously author changes, producing self-documented visual narratives without hand-crafted prompts. BlinkArt’s approach can serve both as a controlled research paradigm and a gaze-reactive art installation in which the artwork evolves precisely where the viewer is not looking.

Dominik Steinmann, Jannis Strecker-Bischoff, Kenan Bektaş, Jessica Laraine Williams, and Simon Mayer. 2026. BlinkArt: Using Change Blindness For AI-Generated Image Modifications in Artworks. In Companion of the 2026 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp Companion ’26), October 11–15, 2026, Shanghai, China. ACM, New York, NY, USA, 6 pages. https://doi.org/10.1145/3798063.3837172