Ceramics Catalog: The Imperfect Collection
Ceramics rarely come out perfectly uniform. The clay settles during firing, the glaze runs to one side, or the handle is attached slightly askew. Usually, this is labeled a defect. But I have always felt that it is precisely in these imperfections that the object becomes alive. It shows that it was made by hand, not by a machine.
«KRIVO» is an exploration of the aesthetics of imperfection through a generative neural network.
I trained a Stable Diffusion XL model (via LoRA) on a collection of ugly ceramics: crooked mugs, skewed vases, plates with wavy edges, and teapots with deformed spouts. The goal wasn’t to generate beautiful pottery, but to teach the model to find beauty in imperfection. I then used it to generate a catalog of objects that never existed but could grace a museum of imperfection.
The project argues that design is not obligated to be flawless. An error is a language. Curvature is the author’s signature, not a defect.
Source images for training:
