Structure every prompt in this order: Subject → Environment → Lighting → Camera/Technical → Style/Post-processing

Portrait of a woman in her 30s, freckled skin, wind-swept auburn hair,
wearing an oversized cream wool sweater, contemplative expression,
standing by a rain-streaked window | soft overcast window light from
camera left, low contrast, gentle shadow falloff | shot on 85mm lens
at f/1.8, eye level, shallow depth of field with soft bokeh | muted
color palette, filmic grain, inspired by Peter Lindbergh, editorial
black-and-white photography aesthetic

Never write vague descriptors ("nice lighting," "blurry background"). Always write technical equivalents ("soft window light, low contrast," "shallow depth of field at f/1.8, creamy bokeh").

Progress:

• Step 1: Intake — identify subject, use case, target platform, and any style references
• Step 2: Analyze references — extract lighting direction, color palette, composition, photographer influences
• Step 3: Construct — build the five-layer prompt (subject, environment, lighting, technical, style)
• Step 4: Optimize — add platform-specific syntax, negative prompts, remove ambiguity
• Step 5: Iterate — refine based on output, document what worked

Layer Details

1. Subject: who/what, specific attributes, pose, expression, materials/textures, scale
2. Environment: location type, background treatment (sharp/blurred/minimal), atmosphere, time of day
3. Lighting: source (natural/artificial), direction (front/side/back/Rembrandt/butterfly/split), quality (hard/soft/diffused), color temperature
4. Technical/Camera: perspective (eye-level/low/high/bird's eye), focal length effect (wide/telephoto/standard), depth of field, exposure style (high key/low key/silhouette)
5. Style: genre (portrait/fashion/product/editorial), era, post-processing (film emulation, color grade, grain), photographer reference

Platform Syntax Cheatsheet

• Midjourney: append --ar 4:5 --v 6 --style raw; use :: multi-prompt weighting for emphasis
• DALL-E: pure natural language, no parameter flags; describe style mixing in prose
• Stable Diffusion: use (term:1.3) weighting syntax; separate negative prompt field
• Flux: favor long, detailed, photorealistic natural-language paragraphs over keyword lists

Example 1 — Product Photography Input: "Luxury watch product shot for e-commerce hero image" Output:

Luxury stainless steel chronograph watch, brushed metal case with
polished bezel, black leather strap, positioned at 3/4 angle on a
dark slate surface | large overhead softbox creating soft gradient
highlight across the case, two strip lights camera left and right
for edge definition and dial legibility, deep charcoal seamless
background | shot on 100mm macro lens at f/8, focus stacked for
full sharpness on case and dial | clean, high-contrast commercial
post-processing, subtle reflection beneath watch, premium editorial
advertising quality, --ar 1:1 --v 6

Example 2 — Environmental Portrait Input: "Chef in his kitchen, authentic documentary feel" Output:

Middle-aged chef with salt-and-pepper beard, white double-breasted
jacket dusted with flour, mid-motion plating a dish, focused
expression, in a professional stainless steel kitchen with hanging
copper pots blurred in background | natural late-afternoon light
streaming through a high side window, warm and directional, soft
shadow falloff | shot on 35mm lens at f/2.2, eye level, shallow-to-
moderate depth of field keeping chef sharp and kitchen softly
legible | warm color palette, slight grain, candid unretouched
documentary style inspired by Peter Turnley

Example 3 — Negative Prompt Addition Input: Same watch prompt for Stable Diffusion Output: Add negative prompt field:

Negative prompt: blurry, low resolution, watermark, text, distorted
proportions, extra hands, plastic-looking material, overexposed
highlights, cluttered background