There is no single sentence that produces the best image in every model. The most dependable prompt formula is a flexible creative brief: it names the output’s purpose, establishes the subject, describes spatial relationships, defines the visual treatment, and protects the details that cannot change.

Use this seven-part structure as a checklist rather than a rigid incantation.

The seven-part formula

1. Purpose and format

State where the image will be used and how the frame should be oriented.

Examples: “square marketplace primary image,” “vertical editorial portrait,” “landscape website hero with title space,” or “focused edit of the supplied photograph.”

Purpose influences composition. A website hero needs usable negative space; a product-detail tile needs a closer crop.

2. Main subject

Name the image’s owner in concrete terms. Include the action when one matters.

One translucent amber-glass table lamp with a wide pleated linen shade, switched on.

For a real product, include factual construction. For a concept, leave low-priority features open.

3. Setting and relationships

Explain where the subject is and how supporting elements relate to it.

On a narrow walnut console against a warm white plaster wall, with one closed book behind the base.

“Behind,” “beside,” “foreground,” and “in the distance” reduce the ambiguity of a keyword list.

4. Composition and camera

Define distance, viewpoint, placement, and depth.

Medium eye-level three-quarter view, lamp on the right third, open wall on the left, straight verticals, background softly receding.

Use camera terminology when it communicates a visible effect. Avoid technical values that do not serve the brief.

5. Light, color, and mood

Describe where the light comes from and how it behaves, then name a restrained color relationship.

Soft blue-hour window light from the left balanced by the lamp’s warm glow, low contrast, amber, walnut, cream, and muted blue palette, quiet evening mood.

Mood should result from visible decisions instead of replacing them.

6. Medium, material, and finish

State whether the output is a photograph, illustration, 3D scene, collage, or other medium. Describe critical textures.

Realistic interior editorial photograph with accurate translucent glass, fine linen weave, subtle plaster texture, and natural contact shadows.

7. Constraints

Protect important details and exclude likely distractions.

Full lamp visible, no people, no additional lights, no readable text, no invented logo, no harsh reflections.

Constraints should respond to the task. Do not paste the same generic negative list everywhere.

Combine the formula

The complete prompt becomes:

Landscape website hero with simple headline space. One translucent amber-glass table lamp with a wide pleated linen shade, switched on, placed on a narrow walnut console against a warm white plaster wall, with one closed book behind the base. Medium eye-level three-quarter view, lamp on the right third, open wall on the left, straight verticals, background softly receding. Soft blue-hour window light from the left balanced by the lamp’s warm glow, low contrast, amber, walnut, cream, and muted blue palette, quiet evening mood. Realistic interior editorial photograph with accurate translucent glass, fine linen weave, subtle plaster texture, and natural contact shadows. Full lamp visible, no people, no additional lights, no readable text, no invented logo, no harsh reflections.

This is detailed, but every sentence controls an identifiable decision.

Adapt the formula by task

Illustration

Replace camera and material details with shape, line, layering, and surface direction:

Horizontal editorial illustration about balancing speed and care, one runner tying a loose shoelace at the center of a path that splits into two simple arrows, flat geometric shapes, bold silhouette, limited navy, coral, and cream palette, subtle paper grain, open upper-right area for title, understandable without text, no logos.

Product image

Increase factual product and light details. Reduce story elements. Protect geometry, material, color, labels, quantity, and included accessories.

Thumbnail

Prioritize one subject, contrast, simple background, small-size readability, and exact negative space. Add final text separately when possible.

Image edit

Replace the seven-part scene with a narrow target, requested change, matching requirements, and invariants. Our practical editing prompt guide provides examples.

Use a reference without copying everything

Run a reference through the Image-to-Prompt Generator or analyze it manually with this step-by-step image-to-prompt method. Sort the findings into the seven parts. Then mark each detail as preserve, change, or remove. This step turns automated description into creative direction.

For example, preserve the reference’s low camera angle and broad side light, change the subject and setting, and remove distinctive branding. The formula is a container for those decisions, not an instruction to reproduce the source.

Common mistakes to avoid

Skipping purpose. Without the final use, the model cannot plan useful space or hierarchy.

Describing objects but not relationships. Tell the model what sits in front, behind, left, or right.

Using mood as a substitute for light. “Calm” becomes clearer with soft contrast, a restrained palette, and uncluttered framing.

Creating incompatible instructions. Review the prompt as one scene before generating.

Protecting nothing. Product details, identities, architecture, and exact text require explicit review and invariants.

Frequently Asked Questions

Must I use all seven parts every time?

No. A concept exploration may need only subject, setting, and treatment. Use the complete checklist when format, accuracy, or consistency matters.

Which part has the biggest effect?

It depends on the failure, but subject clarity and composition often matter more than extra style adjectives. Diagnose the current result before expanding the prompt.

Where does a negative prompt fit?

Put essential exclusions in the constraints block. If the interface provides a separate negative field, move relevant failure terms there without changing the creative brief.

Can this formula work with any image model?

The core brief is model-neutral. Adjust syntax, parameters, and conversational wording for the current tool while preserving the visual decisions.