The Structure of a Great Image Prompt: Subject, Style, Lighting, Mood

There is one structure professional prompt engineers reach for again and again when writing image prompts: Subject, Style, Lighting, Mood. Four layers, in that order, that turn a one-liner into something the model can actually compose. Once you internalise it, your hit-rate on first-try images will climb dramatically.

1. Introduction

Most beginner prompts read like a single sentence — "a girl in a forest" — and the result is whatever the model assumes you meant. Professional prompts read like a tiny brief. They tell the model exactly who or what is being depicted, in which visual language, lit in a specific way, with a particular emotional tone. This four-layer structure is not a rigid template; it is a checklist that ensures you never accidentally leave the most important decisions up to the model.

In this tutorial we will break down each layer, see what happens when one is missing, and assemble a complete prompt step by step.

2. The Concept Explained

Every image you have ever admired — a film still, a magazine cover, a children's book illustration — exists at the intersection of four layered decisions. The artist chose what to depict (subject), how to render it (style), under what light (lighting), and with what feeling (mood). When you write an image prompt, you are making the same four decisions on the model's behalf.

Think of it like ordering a coffee. "Coffee, please" leaves five questions unanswered. "Large oat-milk flat white, extra hot, single shot, no sugar" leaves none. Image prompts work the same way — the model has many defaults, but they rarely match what is in your head.

Subject

Name your hero element first. Be concrete. "A woman" is weaker than "an elderly woman with silver hair, wearing a linen apron". Specify count, posture, and the most distinctive features. If multiple subjects appear, mention their relationship — "standing beside", "facing each other", "embracing".

Style

State the medium and the visual language. "Photorealistic", "watercolour illustration", "1990s anime cel", "3D render in the style of Pixar", "oil painting with thick impasto". Style is the single most powerful word group in an image prompt because it determines almost every other texture and edge in the result.

Lighting

Tell the model where light comes from and what it feels like. "Soft window light from the left", "harsh midday sun", "neon backlight", "golden hour rim light". Lighting decisions are what separate snapshots from images that look like they were art-directed.

Mood

Describe the emotional temperature. "Melancholy and quiet", "energetic and joyful", "tense and cinematic", "warm and nostalgic". Mood pulls colour palette, expression, and composition into alignment automatically.

3. The Problem Without This Structure

a chef in a kitchen

4. The Solution

SUBJECT: A middle-aged Italian chef in a crisp white apron,
kneading fresh pasta dough on a flour-dusted wooden table,
flour clinging to his forearms.

STYLE: Photorealistic, shot on a 50mm prime lens,
shallow depth of field, Kodak Portra 400 film aesthetic.

LIGHTING: Warm golden hour sunlight pouring through a
single arched window on the left, soft shadows on the right.

MOOD: Quiet, focused, nostalgic — a slow Sunday morning
in a small family-run trattoria in Bologna.

--ar 3:2 --v 6 --style raw

5. Step-by-Step Breakdown

1. Name the subject specifically. Age, role, posture, distinctive feature. Lead the prompt with this — early words carry the most weight in the diffusion process.
2. Pick a style anchor. One clear medium plus optionally an era or artist reference. Resist the urge to mix three styles in one prompt — the model will compromise.
3. Design the lighting. Direction (from where), quality (soft, harsh, diffused), and time of day. Reference film stocks for instant photographic mood — "Kodak Portra", "CineStill 800T".
4. Set the emotional tone. Two adjectives are usually enough. "Quiet and reverent", "playful and chaotic". Avoid contradictions like "dark, cheerful".
5. Add parameters last. Aspect ratio, model version, style flags. In Midjourney, that means --ar 16:9 --v 6 --style raw. In DALL·E 3, mention aspect ratio in natural language: "in 16:9 landscape format".

Tip: If a generated image disappoints, audit which of the four layers is weakest. Nine times out of ten, lighting or mood is the missing layer — most beginners over-specify subject and forget the other three.

6. Practice Exercises

7. Key Takeaways