>>2748200That's my understanding as well. I'll elaborate a little on the "base" models that are out there (i.e. the things that go into mixes). Understanding these models will help in understanding why people make different mixes.
The "standard" Stable Diffusion (SD) 1.5/2.1 models are very generalist models suitable for your average normie use cases. It knows artists, celebrities, styles, landscapes, objects, you name it. People call it neutered because there were strong nsfw filters that prevented it from being trained on anything but the most vanilla artistic nudity.
NovelAI is a leaked proprietary model that focuses on drawing people. It's less general than SD but also not neutered, and more than capable of generating nsfw content in a variety of styles from anime to western cartoons to borderline photoreal. People who do hyper-real animated styles like NAI based stuff.
AnythingV3 is made basically for anime waifus. Anything else that it can do is pretty much a side effect. As a result, it's extremely good at what it does, although it's extremely specialized. This thing fucking came out of nowhere and is Chinese or some shit, it caught everyone by surprise and is relatively obscure to normies. It's not specifically trained on porn, but there's a lot of porn in there, it's not neutered.
WafuDiffusion is based on Stable Diffusion. As such it has a lot of similarities but it's specialized for anime styles. As with anything that comes from the huggingface website, it's a normie model.
The other models, R34, Gape60, F222, Zack3D, etc. are porn models. They're specifically trained on porn, and anything else they can draw are byproducts of the base model they started with before training or an accident. As such, they're relatively poor at making stuff that looks good. But that's where mixes come in. These models basically serve to add porn capability to the other base models.
