Another somewhat successful embed. Like Krile, this one works mostly on Anime checkpoint, as it was auto-tagged by WD1.4 and trained on NAI - booru tag training will thus give the best results.

It does give some pretty close looks in photorealism though! I may, however, use this lora to make a more realistic data set and train a second version for that.

Second caveat - this dataset of fanart and screenshots didn't really capture the back of her, so her skirt/bustle won't show up, and back shots will be mostly ???? as a result as well - even if the dataset was fully SFW.

Other than that, hope you enjoy!