AI Mineral Expression Test

This project focuses on the AI's ability to use different minerals in sculptures and figurines in creative and aesthetically pleasing ways.

After researching desirable minerals for the project using no methodology in particular I collected a list of minerals from popular craft use cases like turquoise and opal to rarely seen minerals that I could not pronounce correctly like xonotlite and arsendescloizite. Spellcheck doesn't even recognize some of these minerals!

Using the same method of inserting wildcards into the prompt, I took descriptive terms for the minerals (i.e. micaceous, bladed, wheatsheaf, acicular) to see if the AI was able to express these qualities in composition.

I was very pleased with how well the AI was able to compose the sculptures and it also appears that the AI has a very vast database of images taken from scientific, peer-reviewed sources. The AI didn't seem to flinch at being asked to weave whatever I could throw at it into the piece and did it with flair.

This is incredibly meaningful if you are trying to get specific results for things like armor or trim details and it has also shown to be very useful for genning filigrees. You could reasonably use whatever palette of minerals in any combination and weight you so desire to create unique compositions. The scope of materials that the AI can work with needs more exploration.

Time from inception to realization: 30 hours (mostly getting this very squirrely prompt to work consistently)
Tools used: AUTOMATIC1111 WebUI, GIMP
Artist prompts called: Ellen Jewett

Update on this one: I'm circling back to this one and will be demonstrating how to create gens using layering in editing programs and noise maps to build a sculpture, layer by layer, with incredibly flexible pinpoint accuracy.

After discussions with the artist community I will not sell these since an artist is directly called in the prompt.