Most AI images give themselves away in the first second. The skin is too smooth. The light has no direction. Every strand of hair is perfect in a way no camera ever captures. The fix is not a better model — it is a better prompt.
This guide covers the exact camera, lens, film stock, lighting, and imperfection keywords that push any major model toward true photographic realism. Every technique below works in 2026 across Midjourney V7, Flux Pro, Stable Diffusion, and Imagera's built-in generator. Copy the prompt recipes at the end and paste them straight into your tool of choice.
If you want the broader picture on what makes AI images make AI images look real, that companion piece covers composition, post-processing, and model selection. This article goes deep on the prompt layer specifically.
1.Why Most Photorealism Prompts Fail
The default output of every AI image model is optimised for visual appeal, not photographic accuracy. Without explicit instruction, models smooth skin, distribute light evenly, and eliminate the micro-noise that real cameras introduce. The result looks impressive at thumbnail size and unconvincing at full resolution.
The solution is to speak the language of photography. Cameras, lenses, film stocks, and lighting conditions are concepts these models have absorbed from millions of labelled photographs. When you name them precisely, the model shifts into a fundamentally different rendering mode.
There are five layers to get right:
- Camera body and sensor — signals the overall image character
- Focal length and aperture — controls perspective and depth of field
- Film stock or sensor profile — sets colour science and grain structure
- Light source and quality — determines shadow direction and colour temperature
- Imperfection and texture language — prevents the airbrushed default
Hit all five and the result is an image that reads as photographed rather than generated.
2.The Prompt Keyword Table
The table below maps individual keywords to the photographic effect they produce. Use it as a reference when building or diagnosing a prompt.
| Prompt Keyword | Effect on Output |
|---|---|
| Shot on Canon EOS R5 | Triggers full-frame sensor character, high dynamic range |
| Shot on Leica M11 | Adds rangefinder micro-contrast and colour neutrality |
| Shot on Fujifilm GFX 100S | Medium-format depth and tonal gradation |
| 85mm f/1.8 | Classic portrait compression, creamy background blur |
| 35mm f/2.8 | Wider candid feel, more environmental context |
| 50mm f/1.4 | Natural perspective matching human vision |
| 105mm f/2.8 macro | Extreme surface detail, compressed backgrounds |
| Kodak Portra 400 | Warm skin tones, organic grain, lifted shadows |
| Fujifilm Superia X-TRA 400 | Cool-leaning shadows, fine grain, street photography feel |
| Kodak Tri-X 400 | Black-and-white documentary grain |
| Ilford HP5 Plus | Softer monochrome grain, high contrast potential |
| Golden hour | Directional warm sidelight, long shadow casting |
| Overcast diffused daylight | Flat even light, soft shadow edges, natural colour |
| North-facing window light | Cool, directional, classic portrait studio look |
| Rembrandt lighting | Triangle shadow beneath one eye, high drama |
| Practical lamp with warm spill | Intimate interior realism |
| Visible skin pores | Prevents plastic, airbrushed skin |
| Natural subsurface scattering | Adds translucency depth under skin surface |
| Subtle facial asymmetry | Breaks uncanny symmetry of AI defaults |
| Fine film grain | Micro-texture across the image surface |
| Flyaway hairs, stray strands | Eliminates perfect hair rendering |
| Slight motion blur on edges | Adds kinetic life to moving subjects |
| RAW photo, unedited | Signals unprocessed photographic output |
| Chromatic aberration on edges | Lens fringing that reads as optical realism |
| Vignette | Darkened corners typical of wide-aperture lenses |
3.Camera and Lens Keywords: The Fastest Win
The single highest-leverage addition to any prompt is a camera and lens specification. Compare these two prompts:
- Weak: "portrait of a woman in a cafe"
- Strong: "portrait of a woman in a cafe, shot on Fujifilm GFX 100S, 110mm f/2 equivalent, shallow depth of field, natural window light"
The second prompt activates the model's understanding of medium-format optical characteristics. The background blur has the correct quality. The tonal gradation from shadow to highlight is gradual rather than clipped. The skin retains micro-detail because medium-format glass renders it that way.
For portraits, the 85mm focal length is the workhorse. It compresses backgrounds without distorting facial proportions. The 50mm is better for environmental portraits where you want the subject to feel grounded in their surroundings. The 35mm suits street and documentary work where a candid, slightly wide perspective is appropriate.
For product and architecture shots, a longer lens (100–135mm) removes perspective distortion, while a 24mm or wider adds dramatic spatial exaggeration — useful for architecture but rarely for people.
4.Film Stock Keywords: Colour Science and Grain
Film stock references do two things simultaneously: they set a colour palette and they introduce grain structure. Both are powerful photorealism signals.
Kodak Portra 400 is the most widely understood film stock in this context. It produces warm, slightly muted skin tones with lifted shadow values — the look associated with high-end editorial and wedding photography. Add it to any portrait prompt that needs warmth and organic texture.
Fujifilm Superia X-TRA 400 leans cooler. Shadows pick up a slight green-teal cast. The grain is finer and more uniform. It reads as street photography or travel documentary.
Kodak Tri-X 400 is the reference for black-and-white realism. The grain is chunky and directional. Highlights clip sharply. It is the look of mid-century photojournalism and still reads immediately as analogue.
You can combine a film stock with a camera body: "shot on Leica M11, Kodak Portra 400 emulation, 50mm f/2, natural light" gives the model a fully coherent photographic system to simulate.
Learn more about refining specific image types in our guide to how to make AI faces look real.
5.Lighting Keywords That Change Everything
Light quality in photography is defined by two things: source size relative to subject and direction. Large sources produce soft shadows; small sources produce hard shadows. Name both.
For soft, flattering portraiture:
- "Soft north-facing window light" — large, diffused, cool, even
- "Overcast natural daylight" — giant ambient source, minimal shadows
- "Studio beauty dish, frontal" — defined but still soft
For dramatic or editorial work:
- "Rembrandt lighting, single key light at 45 degrees" — triangle shadow pattern under one eye
- "Hard side light, single bare strobe" — deep shadows, strong texture emphasis
- "Backlit golden hour, rim lighting" — glowing edges, silhouette potential
For interior realism:
- "Warm practical lamp at frame left, ambient room light fill" — mixed colour temperature, realistic domestic scene
- "Neon sign spill, urban night" — hard coloured light with street photography grit
The direction word matters as much as the quality word. "Window light" is vague. "Window light entering from camera left, subject at 45 degrees to source" gives the model a specific geometry to render.
For more depth on achieving natural light in portraits, see make AI photos look real.
6.Imperfection Keywords: The Detail Most Prompts Miss
AI models default to perfection. Every pore is smoothed. Every hair falls into place. Every surface is pristine. Real photographs are not like this, and audiences recognise the difference immediately.
Add a dedicated imperfection block to every photorealism prompt:
For human subjects:
visible skin pores, natural subsurface scattering, subtle facial asymmetry, fine peach fuzz, slight eye redness, natural lip texture, flyaway hairs, micro-expression lines
For environments and objects:
surface scratches, worn fabric texture, uneven paint, dust particles in light beam, fingerprint smudges on glass, natural wood grain variation
For the image itself:
fine film grain, slight chromatic aberration on high-contrast edges, subtle lens vignette, natural colour noise in shadows
Pair this with a negative prompt to reinforce the instruction: "cartoon, illustration, plastic skin, waxy, airbrushed, overly smooth, oversaturated, CGI, 3D render"
7.Copy-Paste Prompt Recipes
Use these directly in Imagera, Midjourney, Flux Pro, or Stable Diffusion.
Recipe 1 — Natural light portrait:
Editorial portrait of a woman in her late thirties, shot on Canon EOS R5, 85mm f/1.8, soft north-facing window light, Kodak Portra 400 emulation, visible skin pores, natural lip texture, subtle facial asymmetry, flyaway hairs, shallow depth of field, RAW photo, documentary photography
Recipe 2 — Street documentary:
Candid street portrait of a middle-aged man at a market stall, shot on Leica M10, 35mm f/2, overcast diffused daylight, Fujifilm Superia X-TRA 400, natural skin imperfections, worn jacket texture, fine film grain, slight chromatic aberration, environmental portrait, photojournalism
Recipe 3 — Indoor dramatic portrait:
Close-up portrait of an elderly craftsman in his workshop, single warm practical lamp at frame left, slight ambient fill, shot on Sony A7R V, 50mm f/1.4, Rembrandt lighting, visible pores, deep laugh lines, calloused hand texture, wood shaving detail in foreground, Ilford HP5 Plus, photorealistic, RAW photo
Recipe 4 — Product with hands:
Ceramic coffee mug held in two hands over a reclaimed wood table, morning north light from window, shot on Nikon Z9, 105mm macro, steam visible, subtle finger skin texture, natural nail imperfections, fine grain, warm colour temperature, RAW photo, lifestyle product photography
Recipe 5 — Black and white documentary:
Street photographer at work in a crowded night market, rain on pavement, practical neon light from above, shot on Leica Q3, 28mm f/1.7, Kodak Tri-X 400, heavy film grain, high contrast, motion blur on crowd at edges, environmental candid, documentary photography
Want to run these on a realistic AI image generator without the trial-and-error? Imagera handles the generation and lets you refine with one click. Plans start at $4.99/month — see Imagera pricing.
8.Putting It Together: The Five-Layer Formula
Every strong photorealism prompt follows this structure:
[Subject + scene] + [camera body] + [focal length and aperture] + [light source and quality] + [film stock or sensor profile] + [imperfection block] + [negative prompt]
The order matters. Lead with the subject so the model anchors on the right content. Follow with technical parameters that shape how the subject is rendered. End with quality and imperfection language that overrides the model's default polish.
Shortcutting any of the five layers pushes the output back toward the illustrated, over-smooth default. All five together produce results that hold up at full resolution.
