This model was trained on SD1.5.

I suggest using CFG 5-7 and anything above 15 sampling steps.

Model used: realisticVisionV13_v13