Why AI Portraits Look Plastic and How to Fix Them

You know the look. Skin like polished wax, not a pore in sight, eyes a little too bright, every face the same age of flawless. It's the fastest way to tell an AI portrait from a photograph, and it's not a limitation of the model. It's what the prompt asked for.

The good news: it takes about three changes to fix.

Why AI skin goes plastic

Models learn from captioned images, and the images captioned “beautiful”, “perfect” and “flawless” are mostly retouched: beauty campaigns, filtered selfies, magazine covers. So when a prompt leans on those words, the model reaches for the retouched end of everything it has seen.

Four habits push portraits in that direction:

• Beauty words. Beautiful, perfect, flawless, gorgeous, attractive. Each one nudges toward airbrushing.
• Quality words. 8k, ultra detailed, masterpiece, hyperrealistic. Ironically, these often produce the most artificial results, because they pull toward digital art and heavily processed images.
• Flat, even light. Light from everywhere removes the shadows that reveal texture. Skin looks smooth when nothing rakes across it.
• No medium. Without a camera, lens or film, the model has no reason to add grain, optical softness or imperfection. It defaults to its own clean, synthetic sharpness.

Fix 1: describe the skin you want

Say it directly:

It feels strange to ask for imperfection, but that's what real faces have.

Add age honestly. “A woman” drifts toward twenty-five with filtered skin. “A woman in her fifties” gets lines, character and credibility.

And remove every beauty and quality word. Replace them with decisions instead.

Fix 2: give it grain and a real film

Film grain is the single most powerful anti-plastic ingredient. It breaks up the smooth digital surface and gives the eye something organic to read.

That works for soft, warm, flattering portraits. For black and white with grit:

Even a subtle grain reference changes the whole feel.

Fix 3: use directional light and a real lens

Texture appears where light rakes across a surface. Window light from the side shows pores, stubble and fine hair. Flat front light hides them.

Then name a real lens and aperture. 85mm at f/2 gives natural optical falloff: sharp eyes, softer ears, a background that melts the way real glass melts it. That optical softness is part of what makes a photograph read as real.

Bonus fixes for stubborn cases

• Documentary framing. Words like candid, documentary photograph and unposed pull away from the studio-beauty cluster.
• A little life in the scene. Flyaway hairs, a crease in a collar, an ordinary background. Perfection in the surroundings encourages perfection in the face.
• Halation and softness. A gentle glow around highlights, like real film, softens the too-crisp digital edge.
• Exclude the look. In tools that support negative prompts, list airbrushed, plastic skin, retouched, CGI, 3D render. In tools that don't, write the positive version instead: unretouched.

Before and after

Before:

After:

The second prompt is barely longer. It just replaces every wish with a decision.

When smooth is right

Sometimes you want polished skin: a skincare ad, a fashion cover, a stylised character. That's fine. The point is to choose it. Say beauty retouching, soft glow, clamshell lighting on purpose, rather than getting it by accident every time.