Trained at native 1024px (with 1024x1024, 1024x768, and 768x1024 buckets) on a dataset of about 350 carefully selected images. Covers a variety of potential interior and exterior locations. Activate with `detailed background` in your positive prompt (all images were captioned with this as the last caption, following a list of appropriate specific tags for each, such as `indoors`, `outdoors`, `windows`, `mountain`, and so on and so forth).
Can be useful for different things depending on the desired image and checkpoint used anywhere from 0.1 all the way up to 1.0 strength, so feel free to experiment.