Support me on Ko-fi:
Description
LoRA trained using 85 images, 81 screencaps from cutscenes in Danganronpa Another Episode: Ultra Despair Girls and 4 from official concept art. I always liked her appearance from the in-game cutscenes the most, however there's a lot of inconsistencies across her representation, so I decided to make my own LoRA which takes purely from screencaps and concept art from the game.
I'm honestly very surprised that this came out as well as it did, admittedly there's some inconsistencies when generating her outfit but it seemed to capture her features quite well. Unfortunately due to the limited nature of the dataset some of the features may not match up exactly but it is what it is. I've also noticed some inconsistencies in eye color but this is pretty rare.
Known Issues
If you're having issues with random blushing or crying then just put blushing, crying, tears, etc. in the negatives and it should go away.
Also, if the results aren't matching the expected, make sure you have the underscores included in the tags, as otherwise it will default to its previous conception of what that tag means (ex. including "green hair" instead of "green_hair" will make her hair appear noticeably green).
Conclusion
This is my first LoRA I'm uploading on here, so feel free to leave a comment with any suggestions, or just to tell me you liked the model. I'm still learning, so keep in mind the model may not be the best. Also please share any works you make, it would be nice to see if people decide to use it.
I'll probably upload more models at some point if this is well received, I have one I'm currently working on but it's proving to be more difficult than expected.