Another pal that's difficult due to their "clothing" except this one has some issues with the purple highlights and adding blue fireballs. I'm hoping to add a new version of this LoRA trained on a larger data set featuring anime/2D images. The main goal is to have an adaptive and consistent LoRA that can easily be used for this character and any custom alterations. V1.0 does not currently adapt well to alterations such as body shape or nudity, more testing is required to find any other persistent issues.

THIS LORA REQUIRES A STRENGTH OF 1 FOR ACCURATE GENS. (sorry)

This LoRA was initially trained using a small data set of screenshots from the game. 3D was added to the training data as a tag. As such removing the 3D tag or adding it to negatives can help produce results in an anime/2D style.

As with all my LoRAs I'm just doing whatever feels right until I get a decent result. In no way do I actually know what I'm doing but sometimes they work well.