Loading…
13
My AI portraits kept getting flagged as fake until I added one human flaw
I spent 3 months generating portraits for a client in Portland and every single one got rejected by their review team. The faces were too symmetrical, skin too smooth, eyes too sharp. Last week I started adding small asymmetries on purpose, like one pupil slightly bigger or a crooked smile line, and suddenly they passed review. The irony is the AI can make perfect images but humans read perfection as a lie. We trained ourselves to spot AI art and now we are punishing it for being too good. Is anyone else building flaws into their prompts just to make the work feel believable?
1 comments
Log in to join the discussion
Log In1 Comment
owens.ben24d ago
Did you notice the rejects got worse when you added more detail to the face? I ran into the same wall doing product shots for a small brand last year, and what finally worked for me was shooting at lower res and then adding a tiny bit of blur on purpose. Something about a slightly soft edge reads as a real camera lens instead of a render. I also stopped centering the subject every time, since a slightly off frame made the review team stop looking for AI tells. The weird part is nobody ever told me what their check was, they just kept bouncing files back with no notes. Once I baked in small mistakes like uneven lighting or a stray hair, approvals got way faster.
7