This model is a lycoris for SDXL. It was trained on the sdxxxl_v10 checkpoint.

The model was trained with 18 pictures found online. The pictures were cropped to remove non-subject matter.

The captions for training were of the form,

"a women fisting her ass to the [wrist|knuckles],from behind, [on knees|standing]"

Any additional captions were added at the end and included details that stood out from the picture.

Examples: "shirt,high heels,standing,astroturf,on bed"

Recommended Settings

The model works well with a,

• 0.8-0.9 strength

• 4.5 CFG

• 30 steps

• range of resolutions that fit the scene

• DPM++ 2S a Karras

Other setting may work as well these gave good results.

Strengths

The model responds to different prompts for clothing, high heels, hair color.

The model can change the position and the locations.

The location needs to have the ability for the subject to be bending over or the model does not work well.

The model does close-up's well, "close-up of a women ...".

Weaknesses

The model struggles with the standing pose.

Sometimes it will draw high heel fragments, adding high heels to either the positive or negative prompt can fix it.

As there were no fingers in the training data, hands are nubby when not generated at the correct location.

It is not great at faces, prompting different a hair color helps.

Version 2

A new version will be released if the model can be improved.

The model would benefit from a more diverse selection of training subjects, one subject is over-represented.

The resolution of the training data ranged from approximately 800x800 to 4000x4000.