To optimise best results: Use models from the preset, such as, specifically, and most preferably, Hakura to ensure that you get most out of this LoRA. Make sure to have the LoRA at strength 0.7-0.9 to optimise best results.

Using "Hakura" will allow you to have better results as it would have more accuracy and consistency to the style. A lot of my "go-to" models don't really work with this LoRA.

I noticed models with very distinct attributes, such as any JunkJuice models, will not be affected by my LoRA. I have attempted to try and incorporate it into my favourite model, which is every DucHaiten, and it has no effect.

However, this does not mean EVERY market model doesn't have affect or EVERY SDXL preset model has affect. Ensure the model you are taking practice with can easily be influenced by styled LoRAs with obvious results to see the LoRA takw effect.

For Hakura, better results range 25-30 sampling steps, I typically use 25 since I'm cheap and constantly run out of points. Note that sometimes the best results require you to have less in the prompt.

Original LoRA from ドフィー / Doffy
Original website: https://pixai.art/model/1826965653696882127?utm_source=copy_web