Basic Overview
This lora changes the concept of "public use" in illustrious models towards what's shown in the sample images. Its trained to work with furry and human characters without issues.
I think the lora is over cooked but I need to do more testing with it however if you want to generate images like the samples with different characters it works great for that.
Currently I'm working on an improved version that can support multiple characters and additional angles however I need to retag the training data so please wait a bit for that version.
Trigger Words and Prompting
The following two tags act as basic trigger words to activate the LoRA
public_use, through_wallFrom there depending on what you want the characters to do you should consider using one of the following keywords
sex, after sex, orgasm, anal_fingering, solo, side view,If you like this lora remember to support the real artists who draw art like this! I would rather you pay one of them a couple dollars then have you tip buzz to me.
Taking Requests
I'm learning how to create better LoRA's so if you have requests please let me know! <3
Usage Policy
Also you are not allowed to sell images generated with this LoRA or use it in such a way that you receive profit from the images generated or the LoRA itself.
Description
FAQ
Comments (5)
Thank you! Any men in the training data?
Yes but the ratio is off so you'll need to do a few generations to get it right. Try using the "male" keyword with a decently high cfg.
Maybe try "masterpiece, best quality, very aesthetic, absurdres, newest, public_use, through_wall, 1boy, male, <put tag for character here>, simple background, no text, anal, ejaculation, penis, twitching penis," as a starting point for a prompt?
@SleepingSelune I'll give it a shot and experiment a bit. If I get any good results I'll be sure to post them and link your lora as a resource. :D
Version 1 definitely seems to have some over-baked styles, but I managed to get some generations with male focus :) Thanks I look forward to a v2
@MysticFrostwind Saw the images you posted! The look pretty good but I can see the overbaking for sure. Maybe 3000steps is too many and 2500 with a more balanced dataset will be better.









