About This Model
Latest version uses a completely new dataset focusing on Mirabel's appearance in the actual film. I felt like it was worth revisiting this in an attempt to create a better LoRA for anyone that wanted to have a "more authentic" Mirabel without the pollution included in earlier versions.
As a warning to anyone attempting to use previous versions, the goal of those models were to help me create a 3D style. The source material that they were based on has a lot of stylistic choices that align her appearance with NSFW art. The latest version does not have that same purpose or NSFW focus.
Possible Issues
Latest version (V2) attempts to align Mirabel's appearance with how she is depicted in the film. However, Illustrious seems to struggle with this kind of 3D style. I tried quite a few different models, but none seemed to really get it authentic to the film. I am still doing more research on this.
Description
Dataset size increased, visual fidelity increased depending on model
Better understanding/replication of background environments
More adherence to trigger word; captures character with fewer tags