Suggestion: for every 1.0 CFG scale you want to lower the models strength by .10, the max in my experience before it starts going into higher fail chance is 0.80 strength with 3.0 CFG.

Test: I tagged many of these images in a specific order, which might influence the results.

Feet for two feet, foot for a single foot:

• Solo, 1 girl, feet focus, bare feet, toes, (optional nail color) toenails, (skin tone), (eye details), (optional eyebrows), (expression, looking at viewer or elsewhere), (mouth details), (optional earrings), (if unique, nose shape), (hair color), (hairstyle), (outfit or nude), (position, preferably seated like 'sitting above'), (optional hand action), (objects like couch or floor), (blurred background).

For anything involving two or more people i essentially labeled everything according to the person by “:” so “female: bare feet” “male: licking”, if its 2 females i put “female:” “girl:”

it might have been better to use bare foot alone as that is used in general for ai foot content but i feel we must distinguish the true meaning of foot and feet for ai so it causes less confusion.