Why Do AI Faces Look Plastic? A 10-Term Realistic Portrait Glossary and a Ready-Made Prompt Template

When you write "4K" or "hyperrealistic" in your prompt, the model doesn't produce detail; it irons the skin, and what's left is smooth plastic. The face looks like a mannequin, and viewers notice it at a glance. The solution isn't asking for quality, it's describing the imperfections and camera details that a real photo has. The next four steps are the four terms in the carousel.

Realism hides in the imperfections. If you don't keep the pores and moles in the skin, you can't give a human feel. This term tells the model to leave pores, moles and small blemishes in; it's the first term that breaks the plastic look.

Human skin lets light through. If you don't give the ears and the tip of the nose that warm redness, the character turns into a statue. This term asks for that warm look of light passing through the skin.

Film feel, not pixels. Breaking digital sharpness and adding grain gives your scene the soul of a real camera. In the template, this appears as two parts: the feel of being shot on a 35mm film camera, and subtle film grain.

Step behind the camera: don't ask for quality, describe the angle of the light and the depth of the lens. Writing only "4K, photorealistic" gives a harsh, detail-less result. Rembrandt lighting is a single key light coming from 45 degrees to the side and above; it leaves a small triangle of light on the shadowed cheek. An anamorphic lens is a cinema lens: it gives horizontal light streaks and oval background highlights. Together they add cinematic depth and character.

The four terms in the carousel and their six companions. They're all already in the template below; know what each one does, and if needed, drop one or move it up front.

• natural skin imperfections → natural skin flaws: pores, moles, small blemishes. This is what breaks the plastic look first.
• visible micro-pores → visible pores. Gives the skin its texture in close-ups.
• peach fuzz → the fine hairs on the face. Makes skin look alive in side light.
• subsurface scattering on skin → light passing through the skin; warm redness on the ears and the tip of the nose. Without it, the face looks like a statue.
• shot on 35mm film → the feel of being shot on a 35mm film camera; softens digital sharpness.
• subtle film grain → light film grain. Real camera texture instead of perfect pixels.
• rembrandt lighting → a single key light from 45 degrees to the side and above; a small triangle of light on the shadowed cheek.
• anamorphic lens → cinema lens: horizontal light streaks, oval background highlights.
• shallow depth of field → shallow depth of field: face sharp, background blurred.
• authentic candid texture → the texture of an unposed, natural moment.

These are the words that iron the skin flat. If you have an old prompt, remove these first and put the glossary terms in their place.

The formula is actually very simple: delete the quality adjectives, add physical camera and texture terms. The result portrait in the carousel was generated with this template, without adding a single extra word.

The only place you need to change in the template is [character detail]. Delete the square brackets and what's inside, and write the person in concrete terms in English: age, hair, clothing, what they're doing, where they are. Example: woman in her early thirties with dark hair loosely tied back, looking out of a window in late afternoon light.

The template works in an empty chat, without adding a reference image.

• ChatGPT: chatgpt.com → New chat at the top left → paste the template into the message box and fill in the [character detail] part.
• Gemini: gemini.google.com → new chat → paste it the same way.
• Midjourney: use the template as is.

The "--ar 16:9" at the end is Midjourney syntax. In ChatGPT or Gemini, delete it and write "in a horizontal 16:9 aspect ratio" instead; for Instagram, write "in a vertical 4:5 aspect ratio." Press send; the image appears in the same chat, below your message, and you can click it to download.

The same template produces a different face every time; the style and texture stay similar, but no two results are exactly the same. If there's a detail you don't like, say it in one sentence in the same chat, or generate again.

For a hair salon or beauty salon: a hairstylist checking a cut in the mirror. You can copy and paste it as is; if you're using ChatGPT or Gemini, replace the "--ar 4:5" at the end with "in a vertical 4:5 aspect ratio."

For a photo studio or a freelance photographer: a photographer holding a film camera on a street at dusk.

For a clothing boutique: a boutique owner arranging a sweater on a rack, in window light.

The fixed part of the template never changes; you only write the character detail at the start to fit your business. Below are filled-in versions for a cafe, a real estate office and an auto repair shop; copy one of the code boxes and change the person and the place to fit you.

Cafe (a barista making latte art):

A real estate agent holding keys in the living room of an empty apartment:

A mechanic leaning on an open hood in a workshop. One last reminder: get permission before generating the face of a real customer or employee.