recommended settings:

sampler/scheduler:

• euler a

• 2M SDE/Karras

steps: 20-40

if you are generating images below 1024x1024 I recommend using hires.fix

As for adetailer, this lora is trained using Ultra high resolution images, so output will still be great without, but it is better with adetailer in general.

notes*

it does have soeun built into the training which you can use by just prompting generic tags like, black hair, long hair etc. but I am working on a dedicated character lora which will be finished soon. reuploads or sales made from this model will not be permitted.