This is an AI-driven moving-image project initiated in 2025. By collecting images of various objects intrinsically linked to “time”—such as wrinkles, seashells, fossils, tree rings, and stalagmites—the project constructs discrete datasets to train specific neural networks. Throughout this process, the work attempts to harness the neural networks’ capacity for dataset fitting and the generative properties of latent space to perform interpolation and video synthesis between various distribution points. This technique allows these objects, originally sculpted by time, to re-emerge as different manifestations of “water”: wrinkles flow like ripples, seashells transform into vortices, fossils appear to boil, the cross-sections of tree trunks shimmer with undulations, and stalagmites are reimagined as calcium carbonate fountains…

By leveraging algorithmic power, the work translates these forms of “congealed time” into perceptible fluid states, exploring the potential of artificial intelligence to reconstruct natural temporality within the realm of human visual perception.