This model attempts to improve generation results for the character Morgan from Arknights.

Her physical features are represented reasonably well by the LoRA as far as my testing indicates, but results may be somewhat fragile.

For one, her outfits aren't particularly well represented, so it's hard to get something similar to her in-game skins out.

Additionally, while barely any prompting seems to be necessary in single character settings to get her features to appear well, introducing additional character LoRAs or even just other characters may require additional, more descriptive prompting to stabilize results.

This probably needs to retrained at some point, either with a more carefully selected set of images (although the sample set used to train is already fairly small at a mere 20 images) or better tagging (this might help? especially with precision of the output).