LoRA trained to generate tilt-shift style images that make things look like miniatures/models.

Performs best with city environments, but I did my best to make it diverse.