Trying to understand the performance that never happened

During a live performance, a computer captured shadows cast by the performer and spectators, forming the dataset used to train a Generative Advveresarial Network. The resulting artificial shadow figures appear projected in the space, blending with the real shadows of visitors. An object-detection system attempts to interpret the hybrid forms aloud, while body-tracking algorithms analyze human and artificial motion together. The work becomes a dynamic, recursive loop in which the machine tries—and fails—to understand what it sees, rendering legibility unstable.