Important: Use 1.1! There is no need to use 1.0 anymore, which was trained on a smaller and lower quality data set.
This is the first SDXL LoRA I have trained. It is using the same dataset from the SD 1.5 version, excluding any images less than 1024x1024. While SDXL works well for many things out of the box, it still can't draw the traditional one-piece nurse uniform well, hence I made this LoRA.
This LoRA works great with the base SDXL as well as all the fine tuned models, both realistic and anime ones. See the sample images for the prompts I used. Note it has a strong bias towards Asian nurses (as most of the training data are of those), but you can still draw beautiful Caucasian nurses by adding prompts like "Caucasian, blonde, blue eyes", etc.
Description
The training set for version 1.1 is twice as large as 1.0. I was able to reuse more pictures from the SD1.5 training set. I also excluded blurry and AI scaled images, which tend to produce un-natural eyes and overly sharpened expressions. You can see the images and the quality improvement should be quite obvious.