What is Danbooru tagging and why does AI care about it?

Danbooru is an anime image board where every upload is labeled with detailed, structured tags — character traits, poses, clothing, art style, lighting, and more. Because millions of images are consistently tagged by humans, the dataset became ideal training data for text-to-image models like Stable Diffusion. Models trained on Danbooru-tagged data learn to associate specific tag combinations with visual concepts, which is why anime diffusion prompts often read like tag lists (1girl, white_hair, red_eyes, school_uniform) rather than natural sentences. The tagging convention is essentially a shared visual vocabulary between humans and models.