This LoRA adds side sex with feet to SDXL.
Works best at 0.7 strength. Works best in portrait.
Trained on stockings, socks, high heels, Converse All Stars, sneakers, fishnets, bare feet and leg warmers. And Asian women, African women, Latina women, European women, African men, and European men.
If you want socks or stockings or fishnets by themselves though, you want to add the tag, "no shoes", and/or putting high heels or sneakers in your negs.
If you want a man's face, say, "brunette man muscular chest erection legs", or "bald african man black tshirt dark skinned erection legs" if you don't want a man face, say, "solo focus, 1man is erection and legs,"
If adetail is over writing your man faces with women faces, and you don't like that, use the [SEP] tag in your adetailer prompt. Something like, "Beautiful woman's face, tall nose, thick lips, freckles, sultry, looking at another, mouth open [SEP] handsome man's face, mustache, 5oclock shadow, closed eyes, open mouth"
Description
FAQ
Comments (10)
WHY EVERYTIME XL??? please also for 1.5 i love that
I made one, but it is trash, it is like 1/10 if even that. I'll keep trying. All my attempts are in my LoRA folder for my Patreons for $1.
Why XL ? Because 1.5 is considerably less able to understand complex things (trained on cropped 512x512, images from database simply didn't have full bodies).
XL is tremendously more able. If you don't like the fact that XL models are better, filter your Civitai results to SD 1.5, and continue living in the past.
Whining won't help.
why SDXL? because SD1.5 can't. i thought that was obvious
There is not a single Lora in 1.5 that is able to do this. XL has a higher accuracy rate, in anatomy, fingers and just overall rendering 2 or more persons.
With 1.5 no matter how good the dataset, prompt and training parameters. This position will always turn out to be some frankenstein. Stop living in the past please. XL is clearly the way to move forward and can only get better if the community as a whole moves with it
@damiang youve right thanks for making me this clear. Im working on it and its cleary HOT !
Only way SD1.5 COULD technically do this is if you turned the image 90 degrees and had them up and down and THEN flipped it back. Maybe op could train them with the first flip training data
Adding to the interest of a 1.5 model if possible! This looks great!



















