ChatGPT Images (OpenAI gpt-image-2) prompt guide

How to prompt ChatGPT Images (OpenAI gpt-image-2)

1. Write a clear brief in full sentences, ordered: scene/background, then subject, then key details, then constraints.
2. State the purpose up front (poster, product ad, UI mockup, infographic, thumbnail) so the model picks the right finish and polish.
3. For complex images use short labelled lines (Scene:, Subject:, Text:, Style:, Constraints:) rather than one long paragraph.
4. Put exact on-image text in quotes or ALL CAPS and specify font style, size, colour and position; spell unusual brand words letter by letter.
5. Ask for the aspect ratio in words ('landscape 16:9', 'vertical 9:16') since ChatGPT has no parameter syntax.
6. For photos, say 'photorealistic' and use camera language: lens, framing, lighting direction, film grain.
7. Ask for real texture and imperfections (skin pores, fabric wear) instead of glossy studio polish when realism matters.
8. List invariants explicitly ('keep the logo, layout and colours unchanged') and say 'change only X' when editing an uploaded image.
9. Give exact counts and spatial relationships ('three bottles, left to right, tallest in the middle'); the reasoning step checks these.
10. For a transparent background, ask for a transparent PNG with an isolated subject and no backdrop or scenery.
11. Keep one coherent visual direction; don't stack conflicting styles.
12. Prefer a clean base prompt and refine with single-change follow-ups in the same chat.

The shape of a good prompt

Purpose: [what the image is for, format]
Scene: [background/environment]
Subject: [main subject, pose/action, count]
Details: [materials, lighting, camera/lens or art style, palette]
Text: "[exact words]" — [font style, size, colour, position]
Constraints: [aspect ratio in words; must keep/avoid; transparent background if needed]

Avoid

• Keyword soup and Midjourney-style --parameters (ignored)
• Vague text requests without the exact quoted wording
• Conflicting style directions in one prompt
• Over-cinematic or heavily retouched wording when you want believable realism
• Decorative clutter in diagrams, slides and UI mockups
• Cramming many changes into one edit instead of iterating

Example

Request a cozy coffee shop poster for autumn

Purpose: a printable vertical poster (2:3) promoting a coffee shop's autumn menu.
Scene: a small corner café on a rainy autumn evening, maple leaves on the wet pavement, warm amber light spilling from tall windows.
Subject: a steaming ceramic mug of spiced latte with latte art in the foreground on a wooden window ledge.
Style: flat, vintage screen-print illustration with visible paper grain; palette of burnt orange, mustard yellow, deep teal and cream.
Text: headline "AUTUMN BREWS" in bold cream serif capitals across the top; smaller line "Pumpkin spice • Maple latte • Chai" in a clean sans-serif along the bottom.
Constraints: no people, no extra text or logos, keep the composition uncluttered.