OpenAI does publish a prompting guide for its image models. An earlier version of this sheet said it did not; this one is built from that guide.
- Current model: ChatGPT Images 2.5, released 8 September 2026. In the API it is “gpt-image-2.5-sunburst and gpt-image-2.5-flare”.
- OpenAI's structure for complex requests: “organize the prompt as scene, subject, details, and constraints, using labeled sections.”
- Exclusions go in the prompt. OpenAI: “State exclusions such as unwanted text, logos, or watermarks.”
- Words in the image go in quotes. OpenAI: “Put required wording in quotes and describe its position and typography.”
- Through the Responses API your prompt is rewritten first, and the rewrite comes back in
revised_prompt. - Known limits: precise text placement and layout-sensitive compositions.
Every rule on this sheet is quoted from the vendor's own published documentation, linked where it is used.
The formula
OpenAI, Image prompting guide, read at source 16 Sep 2026: “For complex requests, organize the prompt as scene, subject, details, and constraints, using labeled sections.”
Template:
Scene: [setting, light, time]
Subject: [who or what, and what the image is for]
Details: [style, composition, aspect ratio, exact text in quotes]
Constraints: [what must not change · what to exclude]
Worked example:
Scene: north-facing pottery studio at dusk, window light only.
Subject: documentary photograph of a ceramicist centring clay on a wheel, for a studio poster.
Details: muted earth palette, shallow depth of field, 3:2. A sign on the wall reads "OPEN STUDIO — SATURDAY" in plain sans-serif, top left.
Constraints: no extra text, no logos, no watermark.
Breakdown — what each part does
Scene
Where it happens and how it is lit. Set it first so the rest has somewhere to sit.
Subject
OpenAI: “Name the subject and intended use, such as a product photograph, advertisement, or diagram.”
Details
OpenAI: “Specify the composition, aspect ratio, and important placement constraints.” Any words in the image go here, in quotes.
Constraints
What must not change, and what to leave out. For edits, OpenAI says “say “change only X” and list the details to preserve”.
Does it take a negative prompt?
OpenAI, Image prompting guide, read at source 16 Sep 2026: “State exclusions such as unwanted text, logos, or watermarks.”
Keep exclusions specific and short — "no watermark" rather than a paragraph of worries — and put them in the constraints section so they do not compete with the description.
Label the four parts, put any required words in quotes, and write exclusions inside the prompt. There is no separate box.
What not to do
When you use the image generation tool in OpenAI's Responses API, “You can access the revised prompt in the revised_prompt field of the image generation call”. If an image does not match your words, read that field before rewriting yours. An earlier version of this sheet also said the chat interface hides the rewrite; no source for that was found, so the claim was removed.
revised_prompt field.Free or paid? What the free plan allows, who owns the output, and when you need to pay: free vs paid AI image, video and music tools.
Where this model sits among the others: which models take a negative prompt, and what each vendor publishes as its formula.
Putting readable words inside a picture is covered on text inside images.
The checklist
Before you send it:
revised_prompt?□OpenAI, Image prompting guide · OpenAI, Image generation guide · OpenAI, Introducing ChatGPT Images 2.5 (read in a browser; OpenAI blocks automated fetches) All read at source on 16 Sep 2026.
Model versions and vendor documentation both move. Re-read the source before relying on a specific number.
Every other sheet in this set: all model cheat sheets.