The title says pretty much everything you need to know. It uses Danbooru-style tagging. It was trained on AnyLoRA, and is something of a proof of concept. Good at producing images that look like pictures of paintings.