Could you please expand on this? I thought it was stable diffusion, I contacted stability.ai and they told me they don't support these sites and don't allow nsfw content.
They're not supported lol. I'm not an expert but since you didn't get any good answers: I believe they are using QLora fine-tunes of SD 1.5 and ControlNet. You can stack multiple QLoras together, like one for art style and one for a specific celebrity. Post-processing includes AfterDetailer. https://twitter.com/SwiftOnSecurity/status/16726420878108917... (not porn)
I believe the most popular UI is Automatic1111, but ComfyUI is good too.
Although now I am more confused than ever with these 6 words that when I google seem to not even have their own sites but just some github code. This stuff is getting way more complex than what I thought would be (I thought they were simply using some api from a main site and white labeling it as their own brand)
Confusion in my brain how these pieces fit together:
SD 1.5 is a base model that is used for fine-tuning. QLora is a technique for fine-tuning with less hardware and time. People distribute the resulting QLora files, which you can load alongside the base model. A1111 and ComfyUI are tools for coordinating the inputs, models, outputs, and post-processing. ControlNet and AfterDetailer are specific projects that would plug in to A111.
thank you brother! I appreciate it. So basically it seems that civit is already using this bundle of tools somehow. Hehe, how ignorant am I.
So basically where would a person run all these tools? like what is the set up? (background: all my experience so far online is running wordpress blogs on shared hosting, so this is a big leap :D