Was inspired by that noodle art trend some years ago.

This model was trained on 25 images generated with dalle3 as proof of concept and I think it works pretty well between 0,8 and 1.