These are LoRA adapter models that can be added to Stable Diffusion models to produce better quality images around a certain theme like beautiful scenery, amazing architecture, etc. While the stock Stable Diffusion model can produce these subjects given the right text prompt, these LoRAs vastly improve the quality of images (of a particular theme depending on the model) because they were trained images scrapped from various image Subreddits and filtered for only the the highest quality ones.