There are two LoRAs included here. I usually use some combination of the two adding up to 0.9+. Neither are heavily trained, so they offer flexibility in characters, location, stlye, etc. I've noticed that other LoRAs fail to preserve the context as well at high weights when dealing with complex prompts. Neither work all that well on their own, so I recommend using a combination of the two.
"Sucking a man's dick, sucking a man's cock" usually gets the ball rolling.
Description
RealisticVision2
FAQ
Comments (9)
Is it my impression (by reading your prompts), or did you merge these two LORAs into one in some pictures?
I found that for some seeds, one works better than the other. For other seeds a combination of the two produces the best results. It is really seed/CFG/prompt dependent.
@jjjdddiop948 Oh no, what I mean was something like this one from the Daenerys picture "<lora:bl0wj0b5-000030:0.9>". It looks different from the other two.
@Vrane Ahh I see. I had about 10 versions going. Narrowed it down to the best 2 and renamed them "-c" and "-m"
Dude, nice LoRA but depicting real celebrities in these kinds of acts is a TOS violation and with good reason. You might want to remix your samples ASAP.
Good call. I guess I saw all the person-specific LoRAs on this site and assumed that was okay. Thanks for the heads up.
which file do you use?
I usually use a combo of both. They have different strengths and weaknesses. It's a pretty iterative process to get good images with these. I wanted to make a LoRA that wasn't just the same pose every time, but it means that it can take some trial and error to get what you want.
I am looking for a LoRA that is trained specifically on blowjobs on guys with their pants on- and the girl taking the penis out of the open zipper of the guy wearing pants. Can you do this, please and thanks?