The Core Formula: Subject, Style, Setting

A Hundred Prompts, One Problem: They Start and End With the Subject

New image prompters typically write one kind of prompt: a subject description. "A robot in a city." "A forest at night." "A cup of coffee." The subject is there, but the image that comes back could be anything — a photograph or an illustration, daytime or moonlit, photorealistic or stylized, intimate or sweeping. The model had to guess everything except the noun.

Experienced image prompters write three things: what the image is of (subject), what it looks like in terms of visual treatment (style), and where it exists (setting). Together these three elements form the core formula that prevents the model from having to make more than a handful of decisions for you. Every technique in this track layers on top of this foundation. Get the formula right first, and every additional modifier becomes more precise because it already has a solid base to refine.

Subject: The Anchor of Every Image Prompt

The subject is the most important element in any image prompt because it defines what the model is building the image around. But "subject" does not mean just a label — it means a specific visual entity. The difference between "a dog" and "a golden retriever puppy mid-leap, ears flying, mouth open in joy" is not creativity — it is specificity. Both are subjects. Only one gives the model enough visual information to produce a consistent result.

Subject description works best when it includes:

• Physical characteristics: species, age, build, coloring, notable features
• Pose and body position: standing, seated, in motion, facing direction
• Action or state: what the subject is doing or experiencing at the moment of the image
• Expression or demeanor: particularly important for human and animal subjects
• Clothing or props: if relevant, described specifically rather than generically

The level of subject detail you need scales with how specific your vision is. If you genuinely do not care whether the dog is a retriever or a terrier, the label is sufficient. If you have a specific image in mind, the description needs to match it.

Style: The Visual Language You Want

Style is the element most beginners either skip entirely or describe too vaguely. "Realistic" tells the model very little — realistic by what standard, in what era, for what medium? "Film photography, 1970s Kodachrome palette, natural grain" is realistic in a specific and reproducible way.

Style specification has three layers that work together:

• Medium: photograph, oil painting, watercolor, 3D render, ink illustration, digital concept art, charcoal sketch. The medium determines the base visual texture and rendering style.
• Era or aesthetic movement: Art Deco, Bauhaus, 1980s synthwave, Victorian botanical illustration, brutalist architecture photography, contemporary editorial. These reference clusters of visual decisions the model learned from training data.
• Technical quality markers: cinematic, high-resolution, sharp detail, moody, low-key, editorial. These steer the rendering quality and visual approach within the chosen medium.

Setting: The World the Subject Inhabits

Setting provides environmental context that shapes everything about how the image is rendered — the default lighting, the depth of the composition, the color temperature, and the emotional tone. "A person sitting" and "a person sitting in a cramped Tokyo ramen shop at 11 pm, steam rising, red lantern light" are the same subject in two completely different images.

Effective setting descriptions specify:

• Location type: indoor or outdoor, natural or built environment, general category (forest, office, beach, alley)
• Time of day: morning, midday, dusk, night — this implicitly sets the light quality and color temperature
• Season and weather: winter snow, summer haze, rainy afternoon, autumn fog
• Specific environmental details: the props, textures, and contextual elements that make the setting feel real rather than generic

Before and After: Applying the Core Formula

Subject only: "A chef in a kitchen."

Subject + style: "A chef in a kitchen. Documentary food photography, available light, grain."

Subject + style + setting: "A chef in his 50s, weathered hands plating a dish with focused intensity. Documentary food photography, available light, slight grain. A small Michelin-starred kitchen at 9 pm — stainless steel surfaces, dim overhead light, steam rising from a pan in the background, other cooks blurred in motion."

Each layer of the formula does a specific job. The subject tells the model what to build. The style tells it what the output should look and feel like. The setting grounds the subject in a world that changes the light, the mood, and the composition. The final version is not longer for the sake of length — every added element controls a visual decision the model would otherwise make arbitrarily.

The Order Matters Less Than the Completeness

The three elements can appear in any order in your prompt. Some prompters lead with style (useful when the visual treatment is the most important thing), some lead with setting (useful when the environment shapes everything), and most lead with subject. What matters is that all three are present and specific enough to do their job.

For DALL-E 3 and Imagen 3, natural language flows best — write it like you are describing the scene to a photographer. For Midjourney, a more compact comma-separated structure also works well. Flux.1 responds best to natural language with detailed layering. The core formula applies across all models; the sentence structure adapts to each.

When to Add Beyond the Core Three

The core formula is your baseline — not your ceiling. Once the subject, style, and setting are solid, each additional dimension (lighting, camera, color) adds precision to a specific visual aspect. The lessons that follow take each dimension individually so you can learn to control them deliberately. The pattern to internalize now: start with the core three, then add modifiers one dimension at a time until the output matches your vision.

Your First Template

Here is the simplest version of a reusable image prompt template built on the core formula:

[SUBJECT description: who/what, appearance, pose, action] + [STYLE: medium, aesthetic, quality markers] + [SETTING: location type, time, environment details]

Fill in the brackets for any visual concept you want to generate. This template alone will produce substantially more consistent results than a subject-only prompt. Add the vocabulary from later lessons — lighting terms, camera specifications, color palette — by appending them to the end, one at a time, until the output matches what you had in mind. The image prompt library at OnePlaceForAI.com contains hundreds of examples structured around exactly this formula — use it to see how other prompters have applied each element across different styles and subjects.