How do you write AI image prompts for editorial-quality fashion photography lighting?

Write editorial fashion lighting prompts in four ordered parts: [lighting setup] + [direction and quality] + [mood/contrast] + [camera and lens]. Name the setup (Rembrandt, butterfly, window, golden hour, hard monolight), specify angle and shadow behavior, describe falloff, then lock realism with a camera-and-lens line — never use generic phrases like 'good lighting'.

Use this prompt anatomy on every shot — it forces the model to render lit photography instead of generically illuminated AI imagery.

Lighting setup (name it) — pick a named photographic setup the model has seen labeled millions of times: Rembrandt key (45° to subject, triangle of light under the off-side eye), butterfly/Paramount (key directly above lens, symmetrical under-nose shadow), clamshell (key above + fill below, beauty-style), split lighting (90° side key, half face in shadow), loop lighting (subtle nose-shadow loop), hard monolight (single bare strobe, no fill), north-facing window (large soft source, indirect daylight), golden hour backlight (low warm sun behind subject), paparazzi on-camera flash (hard frontal, blown highlights, deep falloff).

Direction and quality — every prompt names angle, height, size, and hardness. Examples that work: "45° Rembrandt key, slightly above eye line, hard-edged shadow triangle on the off-cheek", "butterfly key from directly above, soft fill from below, sharp symmetrical under-chin shadow", "north-facing window light camera-left, large soft source, gentle wraparound, minimal shadow density", "hard side-raked key from 90°, no fill, deep falloff into black on shadow side". Hard vs soft, large vs small, raked vs frontal — say it explicitly.

Mood and contrast — give the model the contrast ratio in words: low-key chiaroscuro with crushed shadows, high-key with lifted blacks and pale skin tones, editorial stark with 4:1 key-to-fill, moody single-source with rim separation, hazy backlit bloom with warm flare. Add color temperature: warm 3200K tungsten key, cool 5600K daylight, mixed 4000K with warm practicals.

Camera and lens (the realism lock) — combining a lighting descriptor with a specific camera/lens reference is the single biggest jump from "AI-looking" to editorial: "shot on Hasselblad H6D, 80mm f/2.8, medium-format depth", "Mamiya RZ67, 110mm, shallow depth, film grain", "35mm Portra 400, slight halation", "Leica M11, 50mm Summilux at f/1.4". Lens choice changes compression — 35mm for environmental editorial, 80–110mm for portrait isolation, 24mm for distorted Vogue-style wides.

Five named setups, ready to paste:

• Butterfly: "Butterfly key light from directly above lens, soft fill card below, sharp symmetrical shadow under nose and chin, glossy beauty rendering, shot on Hasselblad 80mm f/4, white seamless"

• Rembrandt: "45° Rembrandt key from camera-right, raised above eye line, hard-edged triangle of light on left cheek, deep shadow side, 3:1 contrast, Mamiya RZ67 110mm, warm tungsten 3200K"

• Window/natural: "North-facing window light camera-left, large diffuse source, soft wraparound on skin, minimal shadow density, neutral 5600K, shot on Pentax 67, 105mm, Portra 400 film"

• Golden hour: "Golden hour backlight low on horizon, warm rim around hair and shoulders, hazy lens flare, soft frontal fill from reflector, 35mm at f/2, slight halation"

• High-contrast studio: "Single hard monolight, side-raked from 90°, no fill, chiaroscuro falloff into pure black, sharp specular highlights on fabric, editorial stark, Leica M11 50mm at f/4"

Per-shot micro-direction. Editorial posing prompts pair lighting with body specifics — "weight on back leg, jaw tension, gaze angle 15° off-camera, key catches collarbone and rim-lights the jawline". Skin rim-lit, environment falling into shadow is the editorial separation move; bake it into your global rules so every shot inherits it.

Probe before you batch. Generate one full-frame shot — character, garment, set, skin, lighting all in it — at your final prompt. If that holds, run the rest; if the light reads flat or plastic, fix the prompt before spending more credits. Skipping this propagates the same lighting failure across an entire campaign.

Model routing matters for lighting. Different image models render light differently: Nano Banana Pro is the strongest for lighting rendering — soft falloff, skin specularity, directional shadow honesty; GPT-Image-2 is stronger for text/design and aesthetic baseline; Recraft holds skin texture well for portraits. invideo's agent runs all the current image models inside one project, so for editorial fashion you can build the aesthetic baseline on GPT-Image-2, then re-render the same prompt through Nano Banana Pro to lock the lighting quality without leaving the project. Where one model produces something that looks too "AI", render the identical prompt across several models in parallel and pick the strongest.

As Hridaye, invideo's creative director, puts it: "Lighting direction specified as 'single hard warm key light, side-raked' with skin rim-lit and environment falling into shadow produces editorial-grade separation in AI-generated fashion images."

What to avoid. Generic terms like "good lighting", "professional lighting", "cinematic", "beautiful light" — they give the model no decision to make and it defaults to flat ambient. Replace with a named setup, an angle in degrees, a shadow behavior, a contrast ratio, and a camera/lens. Also avoid stacking three contradictory setups in one prompt ("golden hour studio window backlight") — the model averages them into mush.

These are the lighting moves that consistently produce editorial output — what works depends on the garment, skin tone, and the brand's visual code, so probe one shot before committing the run.

Watch some of these to see what works for you:

Lighting direction specified as 'single hard warm key light, side-raked' with skin rim-lit and environment falling into shadow produces editorial-grade separation in AI-generated fashion images.

— Hridaye, invideo's creative director