This model was trained on SD1.5. This is a first attempt, but thought it looks pretty good enough to share.

I suggest using CFG 4-7 and anything above 15 sampling steps.

Model used: realisticVisionV13_v13