Updated: Feb 11, 2024
styleHarry Potter Style/Uniforms XL LoRA
This LoRA was trained on scenes from the Harry Potter movies, plus some extra support images and regression examples. I purposely made sure that the actor faces were not exactly replicated.
The LoRA is also trained on emblem/crest close ups, so it can be used to inpaint them (select "only masked"). See this example.
I suggest you give a try at around 1 weight, but if you need more accuracy you can increase the weight (1.2 or more). Please refer to trigger words in the right column.
For cfg/sampler/steps follow the suggestions of the base model. I tested this on AAM XL and DreamShaper XL Turbo.
Please have fun and consider leaving a ❤️ and a good review.