Sculptures are in part shaped by the materials they are made of. If you attempted to make the exact same human bust using a bronze casting, a wood carving, and an assembly of nuts and bolts, you would end up with three sculptures that look very different. The differences among these sculptures would be due to both the inherent properties of the respective materials and the specific techniques required to shape them.

This work explores to what extent generative AI understands these constraints and/or has learned biases that tie certain materials to certain techniques or styles. The complete collection consists of 20 sets of five images each, where the five images of each set represent the exact same sculpture but made from different materials. All sculptures are representations of the human form, including busts, full-body statues, and groups of people. In total, the collection spans 19 different materials representing different sculpting techniques such as casting, carving, or assembling.

All images were made with StableDiffusion SDXL v1.0 using text-to-image generation. The generated images were subsequently upscaled with Gigapixel AI.

Claus O. Wilke, January 2024

https://clauswilke.art