Prompt Engineering for Photographers: Merging Optics with AI Logic

TL;DR

Stop treating AI generators like search engines and start treating them like camera sensors. Your photography knowledge—f-stops, focal lengths, and lighting ratios—is your biggest unfair advantage. This guide translates physical lens physics into AI tokens, teaching you how to force Midjourney and DALL-E to respect the laws of optics. The secret? Don't just name the setting; describe the visual consequence of that setting.

Introduction: The Photographer's Unfair Advantage

Listen up. You’ve spent years mastering the exposure triangle. You know that an 85mm lens flatters a face and a 16mm lens distorts it for drama. You know exactly why golden hour hits different than high noon.

Most people prompting AI are guessing. They type "cool photo" and hope for the best. You? You actually know how light works.

But here’s the friction: AI doesn't have a sensor. It doesn't have glass elements, a shutter curtain, or a physical aperture. It has a neural network trained on billions of image-text pairs. When you type "f/1.8," the AI doesn't mechanically open a lens; it looks for patterns in its training data labeled "f/1.8."

To master prompt engineering, you need to stop thinking like a technician and start thinking like a translator. You are translating the physics of optics into the logic of tokens. This deep dive will show you exactly how to do that.

The Physics of AI: It’s Not Light, It’s Language

When you press the shutter on your Sony A7IV, physics happens. Photons hit the sensor. When you hit "Enter" on a prompt, semantic association happens.

If you prompt Canon R5, 50mm lens, the AI doesn't simulate the optics of that specific lens. Instead, it retrieves the vibe associated with photos tagged with that gear. It pulls up high-resolution, sharp, generally commercial-looking textures.

The Golden Rule of AI Photography:

The model understands the effect better than the setting.

While f/11 might mean "deep depth of field" to you, the AI often ignores the number. But if you type deep depth of field, everything in focus, hyper-detailed background, the AI understands the visual intent. We need to merge these approaches: use the technical term to ground the aesthetic, and the descriptive term to enforce the physics.

The Translator: Mapping Camera Settings to Tokens

Let’s break down the exposure triangle and translate it into language that Midjourney (MJ) and DALL-E actually respect.

1. Aperture (Depth of Field)

The Mistake: Typing f/1.8 and expecting a perfect portrait blur.

The Fix: Combine the f-stop with the visual description of bokeh.

Wide Open (f/1.2 - f/2.8):
- Keywords: Shallow depth of field, subject isolation, creamy bokeh, blurred background, macro photography.
- Prompt Injection: "Portrait of a welder, focus on eyes, f/1.8, extreme bokeh, dreamlike separation."
Stopped Down (f/8 - f/16):
- Keywords: Deep depth of field, edge-to-edge sharpness, hyper-detailed, everything in focus, architectural photography.
- Prompt Injection: "Cyberpunk city street, f/11, intricate details from foreground to background, sharp focus."

2. Focal Length (Compression & Distortion)

AI models are surprisingly good at mimicking focal length because the visual difference between 16mm and 200mm is drastic in their training data.

Wide Angle (14mm - 24mm):
- Visual Effect: Expands space, distorts edges, makes subjects feel dynamic or small.
- Keywords: Wide angle lens, fisheye, GoPro footage, panoramic, distortion, dynamic perspective.
- Use Case: Action sports, massive landscapes, establishing shots.
Standard (35mm - 50mm):
- Visual Effect: Human eye perspective, honest, documentary.
- Keywords: 35mm street photography, 50mm prime, human eye view, documentary style.
- Use Case: Street photography, environmental portraits.
Telephoto (85mm - 200mm+):
- Visual Effect: Compresses background, flatters faces, brings background elements closer.
- Keywords: Telephoto lens, 85mm portrait, background compression, zoom lens, intimate.
- Use Case: Headshots, wildlife, isolating a subject in a crowd.

3. Shutter Speed (Motion)

Since AI generates static images, shutter speed is purely about motion artifacts.

Fast Shutter (1/1000s+):
- Keywords: Frozen action, suspended in air, crisp details, high speed photography, stop motion.
Slow Shutter (1/30s - Bulb):
- Keywords: Motion blur, long exposure, light trails, ethereal, ghosting, smooth water.
- Pro Tip: For DALL-E 3, you must explicitly say what is blurring. "A subway train speeding by with heavy motion blur, while the commuter stands perfectly still."

Lighting & Composition: The Photographer's Eye

This is where you win. Beginners write "good lighting." You write specific lighting setups.

The Lighting Library

Rembrandt Lighting: Creates the classic triangle of light on the cheek. Moody, dramatic, artistic.
Volumetric Lighting: The "God rays" effect. Essential for atmosphere. Use with haze, fog, or dust.
Rim Light / Backlight: Separates the subject from the background. Adds a halo effect.
Butterfly Lighting: Glamour shot lighting, shadow under the nose.
Practical Lighting: Light sources visible in the frame (lamps, neon signs, streetlights).

Compositional Tokens

Low Angle / Worm’s Eye: Makes the subject look dominant, heroic, or massive.
High Angle / Bird’s Eye: Makes the subject look vulnerable or reveals the layout of a space.
Dutch Angle: Tilted horizon. Creates tension, unease, or chaos.
Rule of Thirds: Often ignored by AI unless you specify off-center composition or negative space.

Film Stocks & Camera Models: Aesthetic Shortcuts

Instead of describing 20 adjectives, use a specific film stock or camera system. These are "macro" tokens that carry a massive amount of weight.

The "Raw" Look

Digital: Shot on Sony A7R IV, Phase One XF, 100MP, ultra-sharp, unreal engine 5 (for texture).
Analog: 35mm film, medium format film, grain, light leaks.

Specific Film Stocks (The Cheat Codes)

Kodak Portra 400: The holy grail for skin tones. Warm, natural, fine grain.
- Prompt: "Portrait of a woman, Kodak Portra 400, natural skin texture."
CineStill 800T: The night king. Cool tones, halation (red glow around lights).
- Prompt: "Tokyo night street, CineStill 800T, halation around neon lights."
Ilford HP5: Gritty, high-contrast black and white.
- Prompt: "Street photography, Ilford HP5, high contrast monochrome."
Kodachrome: Vintage 60s/70s look. Vibrant, punchy reds and yellows.
- Prompt: "1970s muscle car, Kodachrome, vintage aesthetic."
Fujifilm Velvia: Hyper-saturated landscapes.

Platform Specifics: Midjourney vs. DALL-E 3

Midjourney v6

Midjourney is the "Art Director." It cares about style and aesthetics more than strict logic.

Parameter: Use --style raw to reduce the AI's tendency to "beautify" everything. This makes photos look more like photos and less like digital art.
Parameter: Use --stylize (0-1000). Lower values (e.g., --s 50) stick closer to your prompt. Higher values (--s 750) let the AI get creative.
Aspect Ratio: --ar 3:2 (Standard 35mm) or --ar 16:9 (Cinematic).

DALL-E 3

DALL-E is the "Literalist." It listens to conversational instructions.

Strategy: It rewrites your prompts. If you want a specific lens effect, tell it why.
Prompt: "Use a wide-angle lens to exaggerate the distance between the foreground flower and the distant mountain."
Correction: If it messes up, tell it: "You made the background too sharp. Re-generate with a simulated f/1.4 aperture so the background is unrecognizable."

The Workflow: Iterative Prompting

Don't expect the perfect shot in one go. Treat it like a photoshoot. You take a test shot, chimp the screen, adjust settings, and shoot again.

The Base Shot: Establish Subject + Action + Setting.
- Prompt: "A boxer resting in the corner of a ring, dark gym."
The Lighting Pass: Add atmosphere.
- Prompt: "A boxer resting in the corner of a ring, dark gym, volumetric spotlight from above, sweat glistening, atmospheric haze."
The Lens Pass: Define the optics.
- Prompt: "Shot on 85mm lens, f/2.8, a boxer resting in the corner of a ring, dark gym, volumetric spotlight from above, sweat glistening, atmospheric haze, background bokeh."
The Film Pass: Grade the color.
- Prompt: "Shot on 85mm lens, f/2.8, a boxer resting in the corner of a ring, dark gym, volumetric spotlight from above, sweat glistening, atmospheric haze, background bokeh, Kodak Portra 800, film grain, cinematic color grading --ar 16:9 --style raw"

Common Mistakes to Avoid

Keyword Soup: Don't just list "8k, 4k, hd, hq, award winning." It dilutes the prompt. Be specific.
Conflicting Optics: Don't ask for "Macro lens" and "Wide angle landscape" in the same prompt. It confuses the AI.
Ignoring the Negative: In Midjourney, use --no to remove unwanted elements. --no illustration, painting, drawing, cgi helps force photorealism.
Over-prompting: Sometimes Photo of a dog, 35mm is better than a 200-word paragraph. Let the AI fill in the blanks, then steer it.

Conclusion: The Hybrid Future

AI isn't replacing the photographer's eye; it's digitizing it. The technical barriers of cost and gear are gone. You no longer need a $50,000 Phase One camera to get the Phase One look. But you do need the knowledge of what that look is.

Your understanding of light, composition, and optics is the bridge between a generic generation and a masterpiece. Use the vocabulary of the lens. Force the machine to see the world the way you do.

Now, go shoot. (Or type).