Images#Image generation#Photorealistic

Metro morning portrait prompt: skin as optics

A photorealistic metro portrait prompt: skin as color codes and optical events, the environment as an active interferer, five closing bans holding the style.

A candid portrait on a Shanghai metro platform: a young woman in a white shirt dress reads a book while the arriving train blurs behind her and the yellow safety line runs at her feet

ModelsGPT Image 2

The prompt

A candid photograph: a young woman on a Shanghai metro platform during the summer morning commute, authentic daily-life photography, a natural unposed moment.
Framing: half-body, eye level, moderate distance — the arriving train streaks into motion blur behind her, the yellow safety line at her feet.
The person: East Asian woman in her early twenties. Almond eyes with natural double eyelids, slightly elongated corners, gaze lowered, absorbed. A straight, refined nose. Skin fair to light beige (NC10–NC20), subsurface scattering under the cool overhead light, micro specular highlights on cheekbones and nose bridge, a faint trace of foundation powder. Fresh summer makeup, nothing heavy, lightly tinted lip.
Wardrobe and props: a relaxed oversized white shirt dress, collar lightly open, sleeves loosely rolled above the elbow. A structured natural-tan canvas tote on her shoulder, a physical novel tucked under one arm, pages held open at a saved spot. Weight on one hip, head down over the book, entirely absorbed, holding a pocket of stillness inside the morning rush.
Background: white-and-grey tiled metro wall, overhead route signage, the arriving train's motion blur on one side, other commuters blurred and distant. Two or three stray hairs displaced by the train's air push — natural, unplanned imperfection, nothing geometrically symmetrical.
Light: cool overhead fluorescents with a slight green cast, flat frontal fill, warm spill from the platform entrance behind.
Texture: fine ISO 400 film grain in the shadows — photographic noise, not CG smoothness.
Aspect 2:3. No watermark, no text overlay, not cartoon, not digital painting, not illustration, not anime.

Portrait prompts fail by being “beautiful but fake” — skin like a beauty filter, sets like backdrops. This one spends its effort in two places: skin is written as optics, and the environment is written as an active interferer.

  • Skin gets a shade code (NC10–NC20) plus optical vocabulary: subsurface scattering, micro highlights, foundation powder — each points at a renderable effect and beats “fair, delicate skin” by a mile.
  • “Stray hairs displaced by the train’s air push” gives the frame a reason no one could have posed: physics breaks the perfection, and the realism lands.
  • Five closing bans hold the line against cartoon, digital painting, illustration and anime — each escape route blocked.

Details worth keeping

  • “Gaze lowered, entirely absorbed” solves the fake eye-contact problem: people reading never face the lens.
  • The “slight green cast” of the overhead fluorescents is the true color of metro lighting; name it and the scene instantly reads right.
  • ISO 400 grain in the shadows explicitly asks for photographic noise and refuses CG smoothness — models do hear that difference.
  • The original case was posted by @ToroJushiAi on X; the prompt here is curated from the open-source repo freestylefly/awesome-gpt-image-2 (MIT license), and the example image is the case’s own. Source material comes from the open community; rights holders can ask us to take it down at any time.
  • For the four-line minimal version of this realist route, see the AI selfie prompt.