Works best at 0.75 weight.

I expunged all crosses from the dataset before training, the only time this model will draw crosses is if tags like 'hospital' or 'stethoscope' are used in the prompt. To avoid this, just put 'cross' in the negative prompt.

This was mostly trained on very suggestive images with a few nsfw ones thrown in for good measure, so it's very geared to that.

When using this model, put the color of the dress and cap in the prompt, i.e. 'white dress, white cap', it can do basically any color.

SD1 was made for testing purposes only.

All showcase images for pony were made on perfectponyxl_v3.

There are a few extra tags that came about as the result of training:

*Might draw a cross

**Buggy (On Pony only)

'nametag'

'heart on cap'

'holding, syringe'*

'center opening' (top of dress open so the cleavage is visible)

'slit dress'

'holding stethoscope, stethoscope'**

'holding, scalpel'**