I do find that I get more face details if I train a lora using this as the base model. So please test that out on your end for specific characters/styles. So images in sample use character specific loras.

Recommended to Use Adetailer for faces and skin.

I usually use Euler A - Automatic

I recommend using these in positive/negative prompts:

Positive: source_realistic, photograph, score_9, score_8_up, score_7_up, score_6_up, masterpiece, best quality, very detailed,

Negative: anime, sketch, cartoon, drawing, (double teeth), (neutral background), (simple background), makeup, jewelry, tongue, cgi, plastic skin, unsharp, deformed eyes, deformed body, deformed face, two belly buttons, 6 fingers, extra fingers, deformed hand,