This post uses SD v1.5 and Anything v3 models.
Let’s start with some serious beauty.
medium body portrait of warrior, long red haired, wind, epic fur, gold and diamond armour. fineart Portrait photographs with a Hasselblad H3DII. extremely detailed
Negative prompt: blurry, deformed
Steps: 40, Sampler: DDIM, CFG scale: 10, Seed: 138480130, Size: 512x512, Model hash: 2700c435Now the fun part, i.e. the standard SD model. The prompt is identical to the one above.
I assume this difference is because anime model is trained mostly on girls (even cats look like girls, trust me). However, the standard one knows is that warrior is typically a man. At least the training dataset contains mostly such images.
And last but not least. The results of the anime model can be greatly enhanced by some prompt magic:
An extremely detailed illustration of a cute beautiful warrior with castle, frowning, intricate full body armor made of gold and diamond, filigree, anime illustration, red hair, detailed sky, detailed sea, detailed eyes. dramatic lighting, Masterpiece, ((intricate detail)), (young girl), Jewel-like eyes, (floating hair), long hair, smooth skin, dress, focus on the subject, HD wallpaper, UHD image, trending on Pixiv
Negative prompt: (((Blurry Eyes))), (((bad anatomy))), ((disabled body)), ((deformed body)), ((missing finger)), ((mutant hands)), ((more than five fingers)), badly drawn hands, lack of detail, (((Low resolution))), ((bad hands)), ((text)), error, cropped, low-quality image, normal quality, jpeg artifacts, signature, watermark, username, weapon
Steps: 50, Sampler: Euler a, CFG scale: 11, Seed: 1654289218, Size: 960x512, Model hash: 2700c435