Inspired by DreamlandAI's style. If this needs to be removed, please let me know.
This is a NSFW forward model, so it may be hard to generate images with blue or green buzz. This is something I'd like to fix in future versions of this model.
DPM++ 2M Karras
CFG 4-6 has been working
30-50 steps to get something really nice
Be sure to use a clip skip of at least 2!
Strength of 0.5-1
I've been having a lot of luck with Hyphoria and Face Enhancer (Strength:0.15).
This model was trained on images of Kim Possible, Belle, Violet, Elsa, Samus, Hermione, Rapunzel, and Gwen Stacy, and will make good images of them without character LoRAs. For other characters, I recommend sprinkling in your character of choice.
There is also a decent bias towards generating multiple views of the same character, so you may need to prompt away from that if you consistently want a single image.
Here is an example for onsite generation.
Description
Retrained to reduce saturation and overtraining.