I wonder if you could change this to allow a video input via prompts to turn your head in various angels, like registering for facial recognition. Then it might not trigger the distaste for photos AND you could get better data for face placement.
(this is due to how this model learns, maybe a future model will alleviate that)