How to Write Better AI Image Prompts: Three Practical Examples
Genoriq Team · Oct 5, 2026
An effective image prompt gives the model enough direction to make useful visual decisions. It does not need to sound technical or contain every fashionable photography term. When a prompt fails, the problem is often that the important requirement is missing, buried, or contradicted elsewhere in the description.
Think of the prompt as a brief to a visual collaborator. Explain what the viewer should notice, how the scene should be arranged, and what must remain consistent. Then use the first result to identify the next useful revision.
The examples below are original prompt examples, not tested outputs or demonstrations of guaranteed model performance.
Start with five decisions
First, name the subject precisely. “A cup” leaves many choices open; “a matte blue ceramic espresso cup with a curved handle” establishes material, color, scale, and shape. Include details that matter to the brief, while leaving unimportant details flexible.
Second, describe composition. Specify whether the image is a close-up, an overhead arrangement, or a wider environmental scene. If you need room for a heading, say where that empty space belongs. This is usually more useful than asking for a generically beautiful image.
Third, describe light. Soft window light, an overcast afternoon, and a hard studio spotlight produce different visual intentions. Choose one main lighting idea instead of stacking incompatible phrases.
Fourth, describe the medium and materials. A watercolor illustration, a tactile paper sculpture, and a product photograph should be clearly distinguished. Materials such as brushed metal, linen, and glazed ceramic help establish the scene's visual character.
Finally, state the practical constraints. These might include a plain background, no decorative lettering, a visible handle, or a specific object that should not be changed.
Example one: a product image
Vague starting prompt
A beautiful coffee mug, professional photography, amazing detail.
The problem is not a lack of enthusiasm. The prompt leaves the mug, viewpoint, setting, and intended use unspecified.
More useful prompt
A matte blue ceramic coffee mug with a rounded handle on a pale oak breakfast table. Three-quarter view at table height. Soft morning light enters from the left, creating a gentle shadow. A folded cream linen napkin sits behind the mug. Keep the mug as the clear focal point and leave uncluttered space on the right for a headline. Natural product photography, realistic ceramic texture, no text or logos.
This version establishes a coherent scene. It also gives you a review checklist: mug shape, handle visibility, light direction, and usable space. If the napkin attracts too much attention, your next edit can address that single issue.
When the image must represent a real product, upload a suitable reference and use a supported editing mode. A text description alone should not be treated as a reliable substitute for the actual product's appearance.
Example two: an editorial illustration
Vague starting prompt
AI creativity, futuristic, colorful, 3D, cyberpunk, minimalist.
Several directions compete here. Cyberpunk environments often imply dense visual information, while minimalist layouts ask for restraint. The prompt also never defines a central subject.
More useful prompt
An editorial paper sculpture showing a small seed opening into three abstract creative forms: a picture frame, a flowing film ribbon, and a curved musical wave. Warm ivory background with muted cobalt and orange accents. One central composition with generous empty space around it. Soft studio shadows, visible paper fibers, calm and thoughtful mood. No lettering, no interface panels, no logos.
The image now has a visual metaphor and a consistent material language. You can revise the palette or the balance between the three forms without changing the entire concept. That makes the direction easier to develop into a coherent article series.
Example three: a place with atmosphere
Vague starting prompt
A cinematic cabin in a forest, 8K, epic, realistic.
“Cinematic” is open to interpretation. Adding a resolution word to the prompt also does not replace selecting an actual output size in the form.
More useful prompt
A small timber cabin at the edge of a misty pine forest just after sunrise. Wide view from a low footpath, with the cabin placed in the left third of the composition. Warm light glows through one window; the surrounding forest stays cool and subdued. Damp wooden steps, a thin layer of fog between the trees, no people. Quiet environmental photography with restrained contrast and believable natural textures.
This prompt specifies the contrast that creates the mood: warm interior light against a cool forest. If the result becomes too dark, ask for a slightly brighter forest while preserving the warm window, instead of replacing the entire description.
Use references with a clear role
A reference image can communicate identity, composition, color, or style. Tell the model which of these you want to preserve. “Use this image” is less precise than “preserve the mug's shape and blue glaze, but replace the background with a bright kitchen.”
When using several references in a supported workflow, make their roles distinct. One image might define the product and another the setting. Conflicting references can make evaluation difficult, especially when you have not decided which features take priority.
For a small edit, name both the change and the invariants. Ask to change the table surface while retaining the camera position, object proportions, and lighting direction. Inspect the result carefully: an instruction to preserve something is a request, not a pixel-level guarantee.
Revise one meaningful variable at a time
Begin with one image while exploring a direction. Review the result against the brief, then identify the largest obstacle to using it. Is the subject too small? Is the crop wrong? Does a background object compete with the focal point?
Write the next instruction around that obstacle. “Move the mug closer to the camera and keep the right third empty” is easier to assess than “make it more premium.” Preserve the successful parts of the earlier prompt so that each revision teaches you something.
Change output resolution after the composition is working, if a larger result is needed and supported. More pixels cannot fix an incorrect object, an awkward pose, or a confused visual hierarchy.
A reusable prompt structure
Use this starting structure in your own words:
Create [subject] in [environment]. Frame it as [composition and viewpoint]. Use [lighting] and [medium or material language]. Emphasize [important detail]. Preserve [required elements]. Leave [needed space] and avoid [specific unwanted content].
You do not need every field every time. A simple illustration may only need a subject, style, and composition. An editing task may require a much more explicit preservation instruction.
Try the structure in the AI image generator, or explore GPT-Image-2 and Gemini 3 Pro Image. Choose the model and mode that fit your inputs, keep notes on useful prompts, and judge success by whether the image fulfills the brief.