An image prompt is a visual brief: subject, composition, style and the constraints you will inspect. Start with one intended use and a few important requirements. Add details to resolve a specific mismatch rather than assuming a longer prompt is better.
Example: brief a workshop illustration
We need an illustration for a beginner workshop notice: one open laptop on a library desk, with space on the left for separately added event text. Use generic objects and an original composition. Keep exact dates and booking instructions in the page’s text layer.
Create an editorial illustration for a library workshop notice.
Subject: one open laptop on a wooden library desk.
Composition: laptop on the right; uncluttered space on the left
for text that will be added later.
Style: flat shapes, soft green and cream palette, clear silhouettes.
Show a generic screen without readable text or brand marks.This is the complete example brief. The generator receives a concrete visual task, not an “expert artist” role followed by unrelated instructions.
Keep instructions separate from tool controls
Choose an available image size, aspect ratio and output format in your selected tool. A phrase such as “high resolution” in the prompt does not set an API parameter. OpenAI’s image-generation documentation distinguishes prompt content from size, quality, background and format controls.
For an existing image, use the tool’s documented editing workflow and an image you are allowed to supply. Specify the change and what should stay fixed. A text request alone does not make an unavailable editing feature appear.
Inspect a candidate against the brief
| Requirement | Pass check | Revision if it fails |
|---|---|---|
| One laptop | Only one laptop is visible | Remove extra devices |
| Space on the left | Headline fits without covering the subject | Move the laptop farther right |
| Generic screen | No false event text or brand mark | Use a plain abstract screen |
| Readable at small size | Subject remains recognizable in the intended layout | Simplify secondary objects |
OpenAI documents remaining text-rendering and composition limitations. Put required wording in an editor or HTML text layer, then check spelling, contrast and mobile cropping in the actual layout.
Example targeted revision
Keep the laptop, palette and flat illustration style. Move the laptop into the right third and clear the left half of foreground objects.
Change one observed issue, compare the candidates at their intended display size, and keep the version that meets the brief. Before publication, confirm permission for reference material, identifiable people and any brand elements. For moving imagery, continue with a one-scene video brief.
If using Midjourney, consult the Midjourney overview for its workflow and controls rather than copying another generator’s settings.
Composition reference to compare
Download the SVG reference and compare it with the brief. The laptop is on the right, the left side is clear, and the screen has no event text. Use these observable requirements when reviewing your own generation.