
A highly structured JSON prompt for Nano Banana Pro designed to generate an ultra-realistic, imperfect elevator mirror selfie featuring two friends. The prompt focuses heavily on preserving the exact identity of reference faces, simulating smartphone camera flaws (grain, slight blur, lens distortion), and maintaining a casual, unfiltered aesthetic.
{ "photo_type": "Elevator mirror selfie (real iPhone capture, indoor neutral lighting)",
"subjects": { "count": 2,
"faces": {
"description": "Use 2 uploaded reference images, keep identity exact, no smoothing, no reshaping, no facial edits, preserve exact facial structure, skin tone, eyes, lips"
},
"pose_description": "two friends standing close in elevator mirror, right subject holding phone taking selfie, left subject slightly leaning with one leg bent upward and one arm raised",
"expressions": [
"right: excited open smile, candid energy",
"left: playful pout expression"
],
"body_language": [
"relaxed posture",
"casual friendly closeness",
"slight asymmetry",
"not perfectly posed"
]
},
"clothing": { "details": [ "left: red puffer jacket, grey sweatpants, sneakers, shoulder bag", "right: light cropped top, grey joggers, hoodie layer, handbag", "natural fabric folds and creases", "realistic outfit texture" ] },
"lighting": { "type": "elevator indoor lighting",
"setup": [
"top ceiling lights",
"metal wall reflections",
"slight uneven brightness across faces"
],
"effects": [
"soft highlights on forehead and cheeks",
"shadow under chin",
"slight dull meta




















































