Make AI Photos Look Real: The Nano Banana Pro Field Guide

Why your AI photos look fake

Your eye is faster than you think. It clocks a render in a fraction of a second, long before you can explain why. The image is too clean. The light is too even. The skin looks poured from plastic. Nothing is out of place, and that is exactly the problem.

Real photographs are never perfect. A real camera fights physics on every frame: glass bends light unevenly, sensors pick up noise, film leaves grain, bright lights bleed at the edges. Those flaws are not mistakes. They are the signature that tells your brain "a real camera was here."

A filter cannot add that signature back. A filter sits on top of the image. Camera physics has to be built into it from the start.

This guide gives you the system. You name the camera body, the lens, and the film stock so the model borrows their physics. Then you layer in the flaws on purpose. The result reads as photographed, not generated.

It is built for Nano Banana Pro, which runs on Google's Gemini 3 Pro and rewards structured, detailed prompts over short conversational ones. The more specific you get about the gear and the flaws, the more real the output.

The fix: camera physics, not filters

Three things do most of the work:

• The body sets the format and sensor look (full-frame, medium format, cinema).
• The lens sets perspective, compression, and how the background melts.
• The film stock sets color, contrast, and grain.

Name all three by their real-world product names. The model has seen millions of photos shot on a Canon 5D with an 85mm and Portra 400. It knows what that combination looks like. Generic words like "professional camera" give it nothing to borrow from.

Then you stack the flaws. That is where the realism stack comes in.

The 12-layer realism stack

Build every prompt by stacking these twelve layers. You do not need all twelve every time. The seven marked ★ are the ones that kill the AI look, so never skip those.

1. ★ Camera body — Sets the sensor format and base color science. Medium format renders differently than a phone. Naming a real body anchors the whole image to a known look.
2. ★ Lens (focal length + optics) — Controls perspective and compression. An 85mm flatters a face; a 24mm stretches a room. Naming a real lens borrows its specific bokeh and rendering.
3. ★ Film stock or color science — Sets color, contrast, and tonal response. This is the single biggest driver of "mood." Portra is warm and soft; Ektar is punchy and saturated.
4. ★ Aperture and depth-of-field falloff — A wide aperture (f/1.4) throws the background out of focus with a gradual, natural drop-off. AI tends to render fake, uniform blur. Naming the f-stop and asking for "gradual falloff" fixes it.
5. Lighting — Source, direction, and quality. Real light comes from somewhere and casts directional shadows. "Soft window light from camera left" beats "good lighting."
6. ★ Grain and sensor noise — Film grain and digital noise break up the too-clean surface. Ask for visible grain in the shadows. This one alone removes a lot of the plastic feel.
7. ★ Optical flaws — Halation, chromatic aberration, vignetting, and lens flare. The fingerprints real lenses leave behind. Covered in the glossary below.
8. ★ Skin realism — Pores, oil, subsurface texture, and small imperfections. The fastest tell of a fake is airbrushed skin. Covered in its own section.
9. Motion — Motion blur and shutter feel. A slight blur on a moving hand or a passing car says "this was a real moment, not a still render."
10. White balance — Color temperature. Warm tungsten, cool shade, mixed light. Real scenes rarely have perfect neutral white.
11. Composition and framing — Angle, crop, and aspect ratio. Slightly imperfect framing reads as human.
12. Capture context — Location, time of day, weather, and atmosphere. Haze, dust, and ambient reflection ground the subject in a real place.

The seven that matter most: body, lens, film stock, aperture/depth-of-field falloff, grain/noise, optical flaws, and skin realism. Get those right and the image stops looking generated. The other five make it great.

Reference: real camera bodies to name

Body	Type	The look it imparts
Canon EOS 5D Mark IV	Full-frame DSLR	Warm, dependable color and flattering skin tones. A safe default for portraits.
Sony A7 IV	Full-frame mirrorless	Clean, sharp, neutral. Modern and crisp without being clinical.
Nikon D850	Full-frame DSLR	High resolution with punchy contrast. Good for detail and texture.
Fujifilm GFX 100	Medium format	Huge detail, smooth tonal gradation, shallow medium-format rendering. Premium editorial.
Hasselblad 500C/M	Medium format film	Square 6x6 frame, painterly and timeless. Classic high-fashion feel.
Leica M6	35mm rangefinder film	Reportage and documentary look. Honest, street, lived-in.
ARRI Alexa	Cinema camera	Filmic latitude and cinematic color. Movie-still energy.
Contax T2	Point-and-shoot film	Nostalgic, slightly soft, candid 90s/2000s look.
Canon AE-1	35mm SLR film	Vintage warmth and gentle grain. Retro lifestyle.

Reference: real lenses to name

Lens	Optical signature
24mm f/1.4	Wide and environmental. Mild edge distortion. Puts the subject inside a scene.
35mm f/1.4	Natural, near-human perspective. Great for candid and documentary framing.
50mm f/1.2	The "standard." Sees roughly as the eye does. Creamy, neutral bokeh.
85mm f/1.4	The portrait lens. Flattering compression, shallow depth, smooth background melt.
135mm f/2	Strong compression. Isolates the subject and stacks the background close.
100mm macro	Extreme detail and texture. For skin, product, and close work.
70-200mm f/2.8	Telephoto compression with sharp subject separation. Sports and editorial reach.
Helios 44-2 58mm	Vintage swirly bokeh. Dreamy, characterful background spin.

Reference: real film stocks to name

Stock	Color, grain, and contrast
Kodak Portra 400	Warm, soft, natural skin tones. Low contrast, fine grain. The go-to for people.
Kodak Ektar 100	Saturated and vivid with very fine grain. Punchy color for landscapes and product.
Kodak Gold 200	Warm, golden, nostalgic. The classic consumer-film look.
Kodak Tri-X 400	Black and white. Gritty, pronounced grain, high contrast. Documentary classic.
Kodak T-Max 100	Black and white. Very fine grain, smooth tones. Clean and detailed.
Ilford HP5 Plus 400	Black and white. Flexible, classic grain, forgiving contrast.
CineStill 800T	Tungsten-balanced. Red halation glow around lights. Cinematic night look.
Fujifilm Pro 400H	Cooler palette, soft pastel greens, gentle contrast. Airy and editorial.

The optical fingerprints glossary

These are the flaws to prompt for on purpose. Name them directly in the prompt.

• Film grain — The fine texture left by film emulsion. Ask for "visible film grain," heavier in the shadows. Breaks up the too-smooth digital surface.
• Halation — A soft red or orange glow that blooms around bright highlights, strongest on stocks like CineStill 800T. Prompt "red halation around light sources" for night scenes.
• Chromatic aberration — Color fringing along high-contrast edges, where a lens fails to focus every wavelength of light on the same point. Shows as thin red, green, or purple outlines. Prompt "subtle chromatic aberration at high-contrast edges." Keep it subtle; a little sells realism, a lot looks broken.
• Vignetting — A gentle darkening toward the corners of the frame. Prompt "soft natural vignette." Pulls the eye to the center the way a real lens does.
• Lens flare — Light scattering inside the lens when a bright source hits it. Prompt "mild lens flare from the sun" or "anamorphic streak flare" for cinema looks.
• Motion blur — The streak left by movement during the exposure. Prompt "slight motion blur on the moving hand" to add a sense of a real, captured moment.
• Depth-of-field falloff — The gradual transition from sharp focus to soft background, not a flat blur. Prompt "shallow depth of field with gradual, natural falloff."
• Sensor noise — The digital equivalent of grain, visible in shadows and high-ISO shots. Prompt "fine sensor noise in the shadows" for a low-light digital feel.

Skin that reads as real

Plastic skin is the single fastest way an image gives itself away. Real skin is busy. Tell the model that.

Always include, when there is a person in frame:

• Pores and texture — "visible pores, natural skin texture, fine peach-fuzz."
• Oil and sheen — "a faint natural sheen of oil on the forehead, nose, and chin."
• Subsurface scattering — "subtle subsurface scattering in the cheeks and ears," the soft way light passes through real skin.
• Imperfection — "small imperfections left intact: a stray hair, a faint blemish, uneven skin tone, fine lines. Not retouched, not airbrushed, not smoothed."

The instruction to not smooth the skin matters as much as the instruction to add texture. Models default to flawless. You have to ask them to stop.

Six copy-paste editorial prompts

Each prompt is fully built and ready. Copy the whole block, swap the one variable marked at the top, and generate. Change only that variable. Everything else is tuned to deliver the realism stack.

Prompt 1 — Golden-hour editorial portrait

promptCopy
# SWAP ONLY THIS → [SUBJECT]
# Example: [SUBJECT] = "a woman in her late 20s with light freckles and curly auburn hair, wearing an oversized cream wool sweater"

Editorial portrait photograph of [SUBJECT], shot on a Canon EOS 5D Mark IV with an 85mm f/1.4 prime lens, aperture wide open at f/1.4 for a shallow depth of field with smooth, gradual depth-of-field falloff and a creamy out-of-focus background. Captured during golden hour with warm, low-angle natural sunlight raking across the face from camera left, producing soft directional shadows and a gentle rim light along the cheekbone and the edge of the hair. Graded with the color science of Kodak Portra 400: warm and natural skin tones, low contrast, slightly muted greens, and fine, visible film grain across the entire frame. The skin reads as real human skin, with visible pores, fine peach-fuzz texture, a faint natural sheen of oil on the forehead and nose, and subtle subsurface scattering in the cheeks and ears. Small imperfections are left intact: a stray hair, a faint blemish, slightly uneven skin tone, and fine lines, not retouched, not airbrushed, not smoothed. Subtle chromatic aberration and gentle purple fringing where the backlight meets the hair. Mild lens flare from the sun and a soft, natural vignette darkening the corners. Realistic catchlights in the eyes. The background is an out-of-focus park, compressed by the 85mm focal length. This is a real photograph with authentic camera physics, not a render, not CGI, not digital art. 3:2 aspect ratio.

Prompt 2 — Neon night street scene

promptCopy
# SWAP ONLY THIS → [SUBJECT]
# Example: [SUBJECT] = "a man in a rain-soaked black trench coat, mid-30s, looking off-camera"

Cinematic night street photograph of [SUBJECT], standing on a wet city sidewalk after rain, shot on a Sony A7 IV with a 35mm f/1.4 lens at f/1.8 for a shallow depth of field and a soft, gradual depth-of-field falloff. Lit almost entirely by colored neon signage and storefront windows, with mixed tungsten and magenta light sources reflecting off the wet pavement. Graded with the color science of CineStill 800T: tungsten-balanced color, deep cinematic shadows, and a distinctive warm red halation glow blooming around every bright light source and neon sign. Visible film grain throughout, heavier in the shadow areas, plus fine sensor noise in the darkest corners from the low light. The skin reads as real human skin under colored light, with visible pores, natural texture, a faint sheen of moisture and oil, subtle subsurface scattering, and small imperfections left intact and unretouched. Subtle chromatic aberration along the high-contrast edges of the neon signs. A soft, natural vignette frames the scene. Slight motion blur on a passing car in the background streaks the lights into soft trails. Reflections, puddles, and atmospheric haze ground the subject in a real place. This is a real photograph with authentic camera physics, not a render, not CGI, not digital illustration. 4:5 aspect ratio.

Prompt 3 — Black-and-white documentary portrait

promptCopy
# SWAP ONLY THIS → [SUBJECT]
# Example: [SUBJECT] = "an elderly fisherman with a weathered face and a wool cap, looking directly into the lens"

Black-and-white documentary portrait of [SUBJECT], shot on a Leica M6 35mm rangefinder camera with a 50mm f/1.2 lens at f/2 for a shallow depth of field with natural, gradual falloff. Photographed with soft, directional window light from camera right, wrapping the face and leaving the far side in gentle shadow for strong dimensional modeling. Captured on Kodak Tri-X 400 black-and-white film: rich, gritty, clearly visible film grain, high contrast with deep blacks and bright but controlled highlights, and a full range of midtones across the skin. The skin reads as real, weathered human skin with deep pores, fine wrinkles, visible texture, a natural sheen, subtle subsurface detail, and every imperfection left fully intact and unretouched. Subtle vignetting darkens the corners. A faint, honest softness sits in the extreme corners from the vintage glass, with mild edge falloff. Realistic catchlights in the eyes. The background is a quiet, out-of-focus interior, compressed and softened by the open aperture. The framing is slightly imperfect and candid, as if caught in a real moment rather than posed. This is a real photograph on real film with authentic camera physics, not a render, not CGI, not AI art. 4:5 aspect ratio.

Prompt 4 — Medium-format fashion editorial

promptCopy
# SWAP ONLY THIS → [SUBJECT]
# Example: [SUBJECT] = "a model in a structured emerald silk gown, tall and poised, one hand resting on the hip"

High-fashion editorial photograph of [SUBJECT], shot on a Fujifilm GFX 100 medium-format camera with a 110mm f/2 lens at f/2.8, delivering enormous detail, smooth tonal gradation, and the shallow, three-dimensional rendering only medium format gives, with a gradual depth-of-field falloff. Lit with a single large softbox from camera left as the key, plus a subtle hair light for separation, producing soft, sculpted shadows and clean specular highlights. Graded with the color science of Kodak Ektar 100: vivid, saturated color, fine grain, and crisp but flattering contrast. The skin reads as real human skin in high detail, with visible pores, fine texture, a controlled natural sheen on the high points of the face, subtle subsurface scattering, and small imperfections deliberately left intact rather than airbrushed away. Subtle chromatic aberration appears only at the highest-contrast edges. A soft, natural vignette frames the subject. Fabric shows real weave, thread, and the way light falls across folds. The background is a seamless studio backdrop, softly out of focus and compressed. Composed as a magazine cover with intentional negative space. This is a real photograph with authentic camera physics, not a render, not CGI, not digital art. 4:5 aspect ratio.

Prompt 5 — Editorial product still life

promptCopy
# SWAP ONLY THIS → [PRODUCT]
# Example: [PRODUCT] = "an amber glass skincare serum bottle with a matte black dropper cap"

Editorial product photograph of [PRODUCT], shot on a Nikon D850 with a 100mm macro lens at f/5.6 for sharp product detail and a softly out-of-focus background with gradual depth-of-field falloff. Lit with directional window light from camera right plus a subtle fill card on the left, producing soft, natural shadows and a single clean specular highlight that defines the form. Graded with the color science of Kodak Portra 400 for warm, true-to-life color, low contrast, and fine, visible film grain across the frame. The product sits on a natural textured surface, such as raw stone or aged wood, with a few real-world details around it: a faint dust particle, a soft reflection, a small water droplet catching the light. Materials read as real: glass shows genuine refraction and edge highlights, metal shows micro-scratches and brushed texture, and labels show real print and paper grain. Subtle chromatic aberration appears at the brightest edges, with a soft natural vignette toward the corners. Shallow background haze adds depth. The framing is editorial and intentional with clean negative space for a headline. This is a real photograph with authentic camera physics, not a render, not a 3D mockup, not CGI. 4:5 aspect ratio.

Prompt 6 — Cinematic environmental portrait

promptCopy
# SWAP ONLY THIS → [LOCATION]
# Example: [LOCATION] = "a dusty desert highway at dusk with distant mountains"

Cinematic environmental portrait of a lone figure standing small within [LOCATION], shot on an ARRI Alexa cinema camera with a 40mm anamorphic lens, framed wide so the setting dominates and the subject anchors the lower third. Shallow depth of field with a gentle, gradual falloff isolates the figure slightly from the deep background. Lit by available natural light at the edge of day, with warm low sun and cool ambient shadow creating a rich, mixed color temperature. Graded for a filmic cinematic look with deep but detailed shadows, soft highlight rolloff, and the wide latitude of motion-picture film. Fine film grain sits across the entire frame. Atmospheric haze, dust, and ambient particles fill the air and catch the light, adding real depth. A horizontal anamorphic lens flare streaks across the frame from the sun, with subtle chromatic aberration at the bright edges and a soft natural vignette. The skin reads as real human skin with visible pores, texture, a faint sheen, and imperfections left intact, even at this distance. The framing is widescreen and composed like a film still, with the horizon slightly off-center. This is a real photograph with authentic camera physics, not a render, not CGI, not concept art. 2.39:1 cinematic aspect ratio.

How to use this with Nano Banana Pro

Nano Banana Pro rewards structure. It reasons through a detailed prompt before it generates, so the more precisely you stack the layers, the closer you land. Work in this order:

1. Pick the scene. Decide what you are shooting: portrait, product, street, fashion, or environmental.
2. Choose your three anchors. Pull one body, one lens, and one film stock from the reference tables. Match them to the mood you want.
3. Stack the seven core layers. Body, lens, film stock, aperture with gradual falloff, grain or noise, optical flaws, and skin realism. These are non-negotiable.
4. Name the flaws out loud. Add halation, chromatic aberration, vignetting, or flare from the glossary, depending on the scene. Keep each one subtle.
5. Add the skin language any time a person is in frame. Always include the "not retouched, not airbrushed" instruction.
6. Set the aspect ratio. Nano Banana Pro supports 16:9, 9:16, 4:5, 3:2, and 2.39:1 cleanly. Match it to where the image will live.
7. Generate, then iterate one layer at a time. If it still looks too clean, push grain and skin texture first. If the background blur looks fake, restate the f-stop and "gradual falloff." Change one variable per pass so you can see what each layer did.

Start with one of the six prompts above. Swap the single marked variable, generate, and adjust from there. Within a few passes you will have a repeatable recipe you can run for every shot.

That is the whole system. Real bodies, real lenses, real film, and the flaws real cameras can never avoid. Stack them on purpose and your images stop looking generated and start looking photographed.

At Hypd we build these repeatable creative systems so solo founders can produce studio-grade content without a studio. Save this guide, run the prompts, and make it your default.

FAQ

Do I need to own any of these cameras or lenses to use the prompts? No. You are naming them so Nano Banana Pro borrows the look it already associates with that gear. The model has learned what a shot on an 85mm and Portra 400 looks like from millions of real photos. You are pointing it at that knowledge, not using physical equipment.

Why does naming a real film stock work better than saying "film look"? "Film look" is vague, so the model averages everything it has ever seen. A named stock like Kodak Portra 400 or CineStill 800T has a specific, consistent color, contrast, and grain signature. Naming it gives the model one clear target to match instead of a blurry average, which is why the output looks intentional.

My image still looks too clean. What do I fix first? Push the two layers that fight "perfect" the hardest: grain and skin. Raise the film grain and add sensor noise in the shadows, then strengthen the skin language with pores, oil, subsurface texture, and an explicit "not retouched, not airbrushed" instruction. If the background blur looks artificial, restate the aperture and ask for "gradual, natural depth-of-field falloff." Change one layer per pass so you can see exactly what moved the needle.