Four runs of one prompt, the most anything here has had. The room, the top and the pose come back every time. The only thing that comes apart is which side is which.
Holds across runs: the whole set. Flat beige wall with the same scuffs low on the left, dark curtain down the right edge, woven rattan table cutting across the bottom. The black spaghetti-strap tank, the long straight black hair, one hand up in the hair at the temple, the gaze off to the side, and soft indoor light with no source you can point at. Four for four. Three of them also put a small mole near her mouth, which the prompt never asks for.
Varies: the crop, and it drifts away from what was asked — two are the close-up in the prompt, two pull back to half-body. What is on the table changes every time: a patterned phone case, a bare phone, an iced coffee with an earphone case, a white cup. Jewellery arrives uninvited, a ring in all four and a thin gold bracelet in one. And how much hair actually reaches the face, from none at all to covering an eye.
Ignored: left and right — but only on the body. Right hand raised holds in three runs and flips in the fourth. Hair partially covering the right side of her face lands on a different side run to run, and in one run covers nothing. Meanwhile a dark curtain on the right is on the right four times out of four. So it is not that the model cannot tell left from right; it is that handedness on the subject of a selfie is the part it will not hold. If which hand matters, plan to fix it in a follow-up turn — the prompt will not get you there on its own.
