Advanced GPT Prompt for Realistic Selfies

2x2 grid, 16:9, pick 4 clever subjects SELFIE_HONEST_FRAME INPUT = [ SUBJECT / DEVICE(phone|ringlight) / ARM(holding) / SETTING / MIRROR(y|n) ] UNKNOWNS d_subj (very short distance → wide-lens facial distortion) · L⃗_device (phone screen / ring as near source) · arm pose (holding kinematics) · mirror geometry (if mirror selfie) LAWS close projection arm's-length + phone wide lens ⇒ nose/forehead enlarged, ears + jaw receding, mild barrel distortion. A telephoto-flat, undistorted face at selfie framing = FORBIDDEN. holding arm the capture arm exists - extended into frame at the correct foreshortened angle, OR out of frame with a raised-arm shoulder/posture. Selfie framing with both hands busy elsewhere and no visible arm / stand = FORBIDDEN. device source a phone screen / ring is a near source ⇒ soft frontal fill, gentle 1/r² falloff (forehead brighter than chin if held high), and a catchlight that is a SCALED IMAGE of the device (rectangle for a screen, ring for a ring-light), positioned where the device is. mirror selfie text/logos reverse; the phone appears in-hand in the reflection; the phone's flash/lens can bloom bright; handedness is consistent everywhere. device vs ambient outdoors, the sun still governs the far side; the device only fills the near side. one exposure any motion (hair, background) shares one T; a phone selfie is usually short-T (frozen). CONSERVATION one device ⇒ catchlight shape + fill falloff + frontal softness agree; one geometry ⇒ distortion + arm angle + framing agree; a mirror ⇒ all reversal and handedness consistent. OVERDETERMINATION d_subj from { facial_perspective_exaggeration + barrel_distortion + framing } L⃗_device from { catchlight_shape + catchlight_position + fill_falloff_gradient } mirror from { text_reversal + phone_in_hand + flash_bloom + consistent_handedness } FORBIDDEN telephoto-flat face at arm's-length framing → close projection selfie framing with no capture arm / stand → holding arm catchlight shape ≠ the stated device → device source mirror selfie with un-reversed text → mirror geometry device fills the whole outdoor scene evenly → device vs ambient background smeared while subject frozen → single T