Trained with Civitai's LORA feature, uses SDXL base checkpoint model for achieving better compatibility across models. Recommended strength settings 0.7 to 1.0 depending on model usage types. Refiner process may be needed if any.

This model is different than the previous ones due to the previous ones was LyCORIS so we don't put a different kind of file as an update to the existing model ones but rather a new category instead.

Also, you can enable these option below from Settings> Optimizations in Auto1111's SD Webui to get better image quality while drastically reducing VRAM usage to only 5GB of VRAM usage (if your GPU supports FP8 feature):

1. FP8 Weight

2. Cache FP16 Weight for LORA