Suggested strength: 0.55-0.75 should work with every prompt. The LoRa works with other character LoRas really good in my testings.
Suggested prompt: "<lora:selfspread:0.65> raw photo of woman (selfspread), (taking a selfie in a mirror):1.3" than adding description of woman + usual qualitypromptwords.
The given activationword is "selfspread" most of the trained images was with girls spreading legs in front of mirror so this post produces the best images in my opinion.
By adding the word "mirror" generated picture will show the borders of the mirror, personally like it gives it a more realistic look. For different poses i usually use "kneeling" "legs up" "standin" or "squatting."
This LoRa doesnt give good results with Anime models! And was made and tested with realistic models and a single comic model.
Hope someone can use this. If you like share your results and leave a comment. Have fun.
Description
FAQ
Comments (9)
In case anyone is asking:
Does NOT work with anime models. Results are awful horror every time. For anime models just prompt "#insertnamehere# spreading her legs in front of a mirror making a selfie with smartphone" and you get the exact same results like this lora.
Thanks for trying Anime Models i really rarely using them and most of the time i can get poses i like without any loras. Im gonna add this info in the description.
nice brompting bro
Good. We got enough anime models/lora's already, 90% of the site is littered with it, you have to look with a microscope to find something that's tailored for realistic results these days.
@gigsa Agree, not against it but I'd love to see more photorealistic nsfw LORA's instead of the 1000000x waifu shit we see here
Such a good LoRa, so much variations. I only wish a tag could come like face covered or uncovered. I've tried extensive prompting but didn't help.
I'd like to create the one made by Kenny in the gallery with phone next to face instead of covered.
RealisticVision
This is great! However, I can't seem to make the 'legs up' prompt work. Any clues?