This is a concept experiment. I wanted to create a model that draws different outfits of the same person side by side. This model draws a doctor's labcoat looks / surgical outfits looks in the same picture.

You should be able to get good results with minimal prompting. Using the trigger words alone should give you decent results. Adding more words to the prompts could instead mess things up. See sample pictures for examples.

This model works with anime character models very well.

A few known issues:

• The surgical outfits are not always consistent - sometimes the mask is off. Sometimes there is an extra mask. The training data were like that.

• When used with certain checkpoints, there is a tendency to draw a surgical cap on the labcoat side as well. If that happens, try adding "hat" to the negative prompt.

• It was trained with landscape pictures. Don't try it with portrait aspect ratio.