Join our Discord to see more images and upcoming updates!
Face Breast Pussy
A split screen LoRA
This LoRA is an experiment to see how well splitt screen images can work. It will create a portrait on the left side and a pussy or hips on the right side. I tried to train on images from the same scene so they match. You can combine it with my other LoRA to improve the details.
Regional prompter can make splitt screen generation easier but is not necessary if you are using the suggested checkpoints.
Model: Big Lust v1.0 (SDXL)
Also works with Big Lust v1.7
SDXL ≠ Pony You should use SDXL models for generation. Mixing this LoRA with pony models will not give you the same quality.
Training Resolution
1248x832, 1:1, Landscape
Tags
Keep in mind: Tags raise the chance to get what you want. But depending on the amount of training images containing this tags some tags might need more luck than others to work.
Main tag
You dont need the main tags. You need to describe the toy!
fbp, split screen, woman, close up
View
front view, side view
Pussy type
cameltoe pussy, innie pussy, normal pussy, butterfly pussy
Shaving
shaved vulva or (line, triangle) + shaped pubic hair
Pussy
spreading pussy, stretching pussy
Pussy gape
[size] pussy gape - The size of the hole
size: tiny, small, medium, big
Other tags
dick, dildo, fingers, butthole
Example prompts
See promo images
Description
first release
FAQ
Comments (7)
dataset assured the both pictures is for same person (real pictures used for both right and left for same person)?
I have not trained this on AI images if that's the question. Right and left belong to the same person and in most cases from the same or similar scene but every image is from another person so you don't recognise anyone. AI is mixing this together so you can get many combinations that are not existing in the training set. If you describe the scene and also background it's easier to get a matching pair.
@Lucifie good first step, but you can make it reach goal that generated images is take the likeness as from dataset which (left image and face and breast , right image = same person) then it narrow things more to be more realistic in term of likeness related on how the prompt describe the body type (other mean , this body is for this face or character) as possible ?
+ far goal is determine a girl naked lower body (or total body) from feeding the sdxl model to a picture of that character in clothes or half but it can understand the body type (might be ipadapter ..) ..hhaha dreams ?
@amazingbeauty Not sure if I get you. I just described the content of the image with tags and hope the AI understand thats the same person. I just hope the AI understands the concept of generating one image in two shifted segments. If I add a tag "two images of the same person" on every image I dont know if that would help.
You can already create an image of the upper body, then fill the right right image with gray color and inpaint the whole image but with a mask on the gray side only. This will create a pussy but still needs the right pussy tags.
It sounds like he wants an unrealistic goal. That is for the model to capable of accurately predicting if the carpet matches the curtains and moreso constraint the supposedly inherent anatomical relation, i. e. what must the pussy of a given girl look, based on her observable genetic and aesthetic traits. You'll probably get matching skintone, but no, not scientifically-accurate phenotype matching, let alone individual variance.