This model was trained on 55 images of alien landscapes, partly generated dalle3 images.

Works well between 0.5 and 1 with 30-60 steps.