I have perfected the P R O M P T
prompt: masterpiece, highest quality, 1girl, solo, ((nekomata okayu)), hololive, virtual youtuber, purple hair, bangs, purple eyes, cat ears, animal ears. animal ear fluff, cat girl, cat tail, short hair, medium breasts, hoodie, black hoodie, pants, white pants, collar, smug, outside
negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry
@fuggy
I guess the "art quality" is decided the same way as literally everything else.
If you take the entire website, and use every word as a datapoint, then you can get eventually a context from words used no matter how pointless.
You don't need to rate the art, if the word "masterpiece" is appearing under the picture set 342. You just know, that user wants pictures with a property, that is connecting all of them.
I wrote a dumb spam filter in Uni, and it changed the way I see AI forever 
And it didn't even have a single neuron.
@LukeAlmighty@gameliberty.club More to do with training dataset being more general I think, no idea how they decide art quality though since the dataset(danbooru) has no such thing