ChatGPT image prompt examples work best when they describe the subject, setting, composition and finish separately. The templates below are written for ChatGPT Images and are meant to be edited, not treated as magic phrases.
OpenAI’s Images in ChatGPT guide says you can describe a new image in chat, choose an aspect ratio, ask for text or transparency, and make follow-up edits. Its prompting guide recommends clear, specific instructions and iterative refinement. Those two ideas are the basis for every example here.
Docs checked
21 October 2026. ChatGPT Images 2.5 and the current Images interface are described in OpenAI’s help centre.
On this page
A simple image-prompt structure
A useful image prompt can be built from six parts:
- Subject: the person, object or scene that must be present.
- Action: what the subject is doing.
- Setting: where the scene takes place.
- Composition: viewpoint, crop and aspect ratio.
- Visual treatment: photo, illustration, collage or another medium.
- Constraints: details to preserve or avoid, plus any exact text.
Put the non-negotiable parts first. If the result misses one detail, change that detail in a follow-up instead of rewriting the whole request.
Product photo prompt
Use this when you need a clean catalogue-style image without inventing a brand.
Create a studio product photograph of [PRODUCT] on a warm off-white seamless background.
Composition: three-quarter front view, product centred with generous negative space, square aspect ratio.
Lighting: one large softbox from the upper left and a subtle fill from the right.
Materials: show the real texture of [MATERIAL]; keep edges crisp and reflections controlled.
Colour palette: [COLOURS].
Do not add a logo, label, slogan or extra accessory.
For a transparent cut-out, add: “Make the background transparent.” OpenAI documents transparent backgrounds as a supported instruction.
Editorial portrait prompt
Create an editorial portrait of an adult [ROLE] in a bright workspace.
Frame: waist-up portrait, camera at eye level, subject slightly off-centre, vertical 4:5 aspect ratio.
Expression and pose: calm, direct gaze; hands naturally resting on the desk.
Wardrobe: [CLOTHING], no visible logos.
Light: soft daylight from a window, restrained shadow, natural skin texture.
Style: contemporary magazine photography, realistic colour, no beauty-filter finish.
Background: a few relevant tools for [ROLE], softly out of focus.
Do not use the name of a real person as shorthand for a look. Describe the wardrobe, lighting and framing you actually want.
Food photograph prompt
Create a top-down food photograph of [DISH] on a [SURFACE] table.
Arrange the main dish at the centre with [TWO SUPPORTING ITEMS] placed asymmetrically.
Use soft morning window light from the left and natural shadows.
Keep the food believable: realistic portions, small imperfections and visible texture.
Palette: [COLOURS].
Landscape 3:2 aspect ratio. No hands, packaging, logo or text.
If the first result feels too polished, ask for one change: “Keep the same composition, but make the styling less perfect and the light more directional.”
Poster with exact text
ChatGPT Images can follow instructions to add text, but spelling is still worth checking before publication.
Create a vertical event poster for [EVENT].
Exact text, including capitalisation:
"[HEADLINE]"
"[DATE AND TIME]"
"[LOCATION]"
Layout: headline in the upper third, date and location below, one abstract [SHAPE OR OBJECT] in the lower half.
Style: bold geometric print, [COLOUR 1], [COLOUR 2] and black on cream.
Keep all text fully visible and leave a 10% safe margin around every edge.
Do not add any other words, logo or QR code.
Aspect ratio: 4:5.
After generation, zoom in and check every character. If one line is wrong, select that area in the editor and request the exact replacement. OpenAI notes that selections are not always precise and edits may extend beyond the highlighted area, so inspect the surrounding design too.
Flat illustration prompt
Create a flat editorial illustration about [CONCEPT].
Show [MAIN OBJECT] connected to [SECOND OBJECT] by one clear visual action.
Use thick black outlines, simple geometric shapes and a limited palette of [COLOURS] on a warm cream background.
Wide 16:9 composition with the focal object in the centre-left and empty space on the right.
No gradients, photorealism, labels, letters, numbers or interface screenshots.
This format is useful for article covers because it describes a relationship instead of asking for generic “AI art”.
Room redesign prompt with a reference image
Upload a clear photo first, then use:
Redesign this room as a [STYLE] workspace while preserving the existing walls, windows, doors and fixed cabinetry.
Replace only: [MOVABLE ITEMS].
Add: [NEW ITEMS].
Palette: [COLOURS AND MATERIALS].
Viewpoint: keep the same camera position and lens perspective as the reference photo.
Show realistic circulation space and do not enlarge the room.
OpenAI says you can upload an image and describe edits in the conversation. For a local change, name the area directly or use the selection tool. If the reference contains private documents, faces or addresses, remove or cover them before upload.
Logo exploration prompt
Create six black-and-white symbol concepts for a fictional brand called [NAME], which provides [SERVICE].
Concept: combine [IDEA A] and [IDEA B] without using a literal mascot.
Style: simple vector-like geometry, one colour, no gradients or shadows.
Place the six marks in an evenly spaced 3-by-2 grid on white.
Show symbols only. Do not add the company name, mockups or presentation text.
Each symbol must remain legible at 24 pixels.
Treat the grid as exploration, not a finished trademark. Check for similarity to existing marks before adopting one.
How to refine the first result
OpenAI recommends iteration. Make one or two controlled changes at a time:
- “Keep the subject and camera angle. Change only the background to dark green.”
- “Keep the layout. Replace the upper-right object with a brass desk lamp.”
- “Preserve the character’s face and clothing. Widen the crop to 16:9.”
- “Keep every visual element. Correct the headline to: ‘[EXACT TEXT]’.”
- “Remove the selected object and reconstruct the wall behind it.”
The Images editor supports selecting an area, changing the aspect ratio and describing edits without a selection. A request tied to a specific region is less ambiguous than “make it better”.
A final prompt checklist
Before sending a prompt, check that it says:
- what must appear;
- what the subject is doing;
- where the scene happens;
- how the camera or layout should frame it;
- which medium and colours to use;
- which text must be exact;
- what must not be added; and
- which aspect ratio you need.
For OpenAI’s own structure and plan notes, see ChatGPT image prompts; for more playful ideas, try Chat GPT prompts for cool images. The image and video prompting guides collect the rest.
