Realistic Portrait Prompts: The 7-Layer Structure (Gender-Neutral)

10 gender-neutral realistic portrait prompts plus the 7-layer structure that fixes plastic AI skin and dead eyes — tuned for Midjourney V8.1, Flux 2 Pro, and Nano Banana 2 (June 2026).

Realistic portraits fail in two reliable ways: plastic AI skin and dead eyes. Both come from the same root — the prompt never specifies lens, lighting, skin texture, and gaze, so the model falls back on its smoothed-out average face. Below are 10 gender-neutral templates plus the 7-layer structure that produces magazine-grade output on the three models worth using in June 2026: Midjourney V8.1, Flux 2 Pro, and Nano Banana 2.

TL;DR

  • Write four layers every time: subject, skin texture, lighting direction, and lens + f-stop. These four kill most of the AI look on their own.
  • natural skin texture with visible pores plus a named light direction (Rembrandt, loop, split) removes the bulk of “plastic skin.”
  • Pick by model: Flux 2 Pro for the most accurate skin and hands, Midjourney V8.1 for editorial atmosphere, Nano Banana 2 when you need the same face across several shots.
  • The 10 templates below are copy-ready. Swap the bracketed subject details and keep the technical layers intact.

What a high-quality prompt should contain

A stable realistic-portrait prompt has seven layers. The first four are non-negotiable:

  • Subject: age, expression, pose — replace vague adjectives with concrete cues
  • Skin texture: natural skin texture with visible pores — removes most of the AI look
  • Lighting: Rembrandt / loop / butterfly / split — name the direction the light comes from
  • Lens: 50mm / 85mm / 35mm plus an f-stop — controls depth of field and facial compression
  • Background: plain neutral grey / north-facing window / studio seamless — specifics beat “nice background”
  • Film / grade: Kodak Portra 400 / Tri-X / Fuji 400H — a real film stock reads better than “vintage”
  • Aspect: --ar 4:5 portraits, --ar 21:9 cinema, --ar 1:1 social

Which model for realistic portraits (June 2026)

Three models lead for human faces right now. They behave differently, so the same prompt will not land identically on each.

ModelStrongest atEntry price (June 2026)Notes
Flux 2 Pro (Black Forest Labs)Skin pores, hands, anatomy, fabric drapingAPI / per-image (no fixed plan)Most literal; renders in ~15–30s; best technical realism
Midjourney V8.1Cinematic atmosphere, lighting wrap, “looks-finished” outputBasic $10/mo ($8 annual)Default version since June 2026; add --v 8.1 --style raw --s 50 for photoreal
Nano Banana 2 (Gemini 3.1 Flash Image)Character consistency across runs, up to 4KAPI / per-image, scaled by resolution (higher at 4K)Released Feb 2026; accepts multiple reference images to lock a face

Practical rule: roll the Flux 2 Pro and Midjourney V8.1 versions of the same prompt and pick the better frame. Use Nano Banana 2 with a reference image when you need one person to appear in multiple shots without their face drifting.

The --ar and --style flags in the templates below are Midjourney syntax. On Flux 2 Pro and Nano Banana 2, drop the flags and set aspect ratio in the interface instead — the descriptive layers carry over verbatim.

10 copy-ready prompt templates

1. Professional headshot base

Best for: LinkedIn headshot, brand bio

A 30-year-old person, calm confident expression, natural skin texture with visible pores, soft Rembrandt key light from camera-left, plain neutral grey backdrop, 85mm portrait, f/2.8, sharp eyes, half body, --ar 4:5

2. Cinematic portrait

Best for: Editorial, personal-brand visual

A 35-year-old person looking off-frame, brooding expression, anamorphic 35mm cinematic look, dim practical streetlight from behind, magenta neon reflection on cheekbone, shallow depth of field, --ar 21:9

3. Natural-light lifestyle

Best for: Lifestyle brand, blog hero

A person sitting by a north-facing window, soft overcast daylight, slight three-quarter angle, gentle smile, 50mm f/1.8, shot on Kodak Portra 400, warm muted palette, --ar 4:5

4. Genuine laugh close-up

Best for: Ad campaigns, family portraits

Close-up portrait of a person mid-laugh, eyes crinkled, natural light from a soft overcast sky, 85mm f/1.4, ultra-shallow depth of field, slight catchlight in eyes, --ar 1:1

5. B&W editorial

Best for: Magazine cover, fine-art portfolio

Editorial black-and-white portrait of a person, single hard key light from above, deep shadows under jaw, neutral expression, smoke haze in air, 85mm f/2.8, --ar 4:5

6. Golden-hour backlight

Best for: Wedding, romantic shoots

Outdoor backlit portrait of a person, golden hour sun haloing the hair, soft fill from a white bounce, slight lens flare, 85mm f/1.8, warm golden palette, --ar 4:5

7. Documentary portrait

Best for: Feature stories, character studies

Documentary-style portrait of a tradesperson in their workspace, wearing work clothes, focused expression, natural workshop light with single soft window, 35mm f/2, slight film grain, --ar 3:2

8. High-key beauty

Best for: Cosmetic ad, brand endorsement

High-key beauty portrait of a person, plain white seamless background, soft beauty dish above and a reflector below, ultra-clean skin texture, sharp eyelashes, 100mm macro, f/8, --ar 4:5

9. Environmental (library)

Best for: Author / academic profile

Environmental portrait of a person standing in a vintage library, warm tungsten lamp light, dust motes in air, three-quarter pose, calm expression, 35mm f/2.8, --ar 4:5

10. Studio fashion

Best for: Brand campaign, comp card

Studio fashion portrait of a person against a deep teal painted backdrop, large softbox key from camera-right, hair-light from behind, sharp jaw shadow, 85mm f/4, --ar 4:5

Common mistakes

  • Vague adjectives (beautiful, gorgeous) — the model can’t translate them into lighting or pose
  • No lighting direction — output looks flat, plastic, and lit from nowhere
  • Skipping focal length — the default ~50mm enlarges the nose at portrait distance
  • Buzzword spam (8k, ultra-realistic, photorealistic) — adds noise, not detail; on Midjourney V8.1 it often fights --style raw
  • Stacking multiple film stocks — the palettes fight each other

How to push results further

  • For crisp pupils: add sharp eyes, catchlight in eyes
  • For magazine vibe: add editorial portrait, magazine cover quality
  • For consistent skin: in-paint the face separately (Flux 2 Pro inpainting) instead of re-rolling from scratch
  • For a “real person” look: negative-prompt plastic skin, doll-like, airbrushed
  • To keep hair detail: add flyaway hair, individual hair strands
  • On Midjourney V8.1: append --v 8.1 --style raw --s 50 to suppress its built-in stylization and stay literal

FAQ

Q: Why does AI skin look so plastic?

A: The prompt didn’t ask for texture, so the model smooths everything. Add natural skin texture with visible pores, avoid smooth skin, and negative-prompt airbrushed, plastic skin. Flux 2 Pro renders pores most realistically of the three current models.

Q: Why do the eyes always look off?

A: Usually no gaze direction. Add looking directly at camera, looking off-frame to the left, or looking down, plus catchlight in eyes so the pupils read as alive rather than glassy.

Q: Midjourney V8.1, Flux 2 Pro, or Nano Banana 2 for real people?

A: As of June 2026, Flux 2 Pro leads on skin pores, hands, and anatomy; Midjourney V8.1 wins on cinematic atmosphere and “finished” feel; Nano Banana 2 is the one to use when you need the same face across several images, since it can lock a subject from several reference photos. Roll the first two and pick the stronger frame.

Q: How do I avoid the “Instagram face” default?

A: Combine skeletal cues (high cheekbones, soft jawline) with real-person imperfections (subtle freckles, slight asymmetry). Imperfection reads as real; flawless symmetry reads as AI.

Q: Do the --ar flags work on Flux 2 Pro and Nano Banana 2?

A: No. --ar and --style are Midjourney syntax. On Flux 2 Pro and Nano Banana 2, delete the flags and set the aspect ratio in the interface; everything before the flags transfers unchanged.

Tags: #Portrait #Realistic #Image generation #Prompt