Recommendations:
In pixai use: hassaku, moonbean and orangeabyssmix
Sampler: euler a for best results
25 steps or more
7 or 8 in CFG scale
Use detailed indications e.g.: 1 girl, room, standing, looking at viewer etc... otherwise only the face of the character could be shown or deformed.
Compatible with many character loras (it will depend on how the lora was trained, sometimes the style will be applied more, sometimes it will be less noticeable.
I recommend using anylora for image generation, but you can use another model that does not have a very marked style.
Weight from 0.9 to 1, with the trigger word for best results.
That's all, enjoy!
