GPT Image 2.5 prompt — a science diagram written as a teaching brief
List the required labels by name. Anything you leave to the model is a label you now have to check.

| Mode | Model | Quality | Size | Ratio | Credits |
|---|---|---|---|---|---|
| Text to Image | Think | HD | 1K | 3:2 | 24 cr |
Published by OpenAI as an example for this model — not generated on this site. Model: gpt-image-2.5-sunburst. Source: OpenAI The published example ran at 1536x1024, a legal size we do not currently offer; the frame here is the same shape at our nearest size.
Cellular respiration for high-school students: glycolysis, the Krebs cycle, the electron transport chain, with eight molecules named individually in the prompt. This is the first template in the library that moves up to HD, and the reason is small text — a diagram is mostly typography, and typography is what a lower quality tier spends less on.
The prompt, nothing cut
Create a simple biology diagram titled "Cellular Respiration at a Glance" for high school students. Show how glucose turns into energy inside a cell. Include glycolysis, the Krebs cycle, and the electron transport chain. Use arrows to connect the steps, and label the main molecules: glucose, pyruvate, ATP, NADH, FADH2, CO2, O2, and H2O. Make it look like a clean classroom handout or slide, with a white background, simple icons, clear labels, and easy-to-read text. Avoid tiny text, extra decoration, or anything that makes the diagram hard to understand.
This exact prompt, at these exact settings
The mode and all five settings are locked to this template, so there is nothing to line up by hand. Press Run this prompt above and the panel opens filled in — prompt, mode, quality, size and frame — and your first image needs no account. If your plan sits below what this ran at, the line under the panel says so.
This example was published at Think · HD · 1K. A free run is Fast · Standard · 1K, so expect less fine detail and softer small text. The composition is the same; the finish is not.
What is safe to change
Most libraries publish this list and stop, which is why so many copied prompts come back worse than the original: the swappable nouns are the safe half. The three sections under it are the other half — the clauses that are doing the work, the ones that break it if you touch them, and why it is written in this order.
The named audience
"For high school students" sets the depth of the whole thing — which steps get shown, how much jargon survives, how much is simplified. Change it to "for an undergraduate lecture" and the same prompt produces a denser diagram.
The explicit label list
Eight molecules, named. This is the difference between a diagram about cellular respiration and a diagram of it. Every label you do not name is one the model chooses, and one you then have to verify.
The quoted title
"Cellular Respiration at a Glance" in quotes gets rendered as written. Titles are short strings, which is where text rendering is most reliable — worth using.
Three ways to break it
Each one is a real failure with a reason attached. A rule without a reason is not usable.
01Running it at Standard
The official example moves to HD specifically for this class of image, and the reason is legibility. A diagram is typography with pictures attached; at Standard the small labels go soft and the artefact stops functioning as a teaching asset.
02Shipping it without checking the biology
Arrow direction and label placement are the two things that go wrong, and both look completely convincing when wrong. A diagram that misplaces NADH is worse than no diagram, because a student will learn it.
03Asking for a portrait frame
A process with three sequential stages wants horizontal room. Forced into portrait, the model stacks the stages and the arrows start crossing, which is where a clean diagram turns into a tangle.
Inside the Make a classroom diagram with the labels you specify prompt
The framing OpenAI recommends for this class of image is to prompt it like an instructional design brief rather than an illustration request: define the audience, the lesson objective, the visual format, the required labels and the scientific constraints. That is a genuinely different way of writing a prompt and it is worth internalising, because it changes what you spend words on.
Notice what is not in this prompt. No colour palette. No art style. No composition instruction. No mood. The visual direction amounts to "clean classroom handout or slide, white background, simple icons". Everything else is content: who is looking at it, what they should understand afterwards, which stages appear, which eight molecules must be labelled.
That allocation is correct because in a diagram, appearance is subordinate to correctness in a way it is not for any other kind of image. A beautiful diagram with the arrows the wrong way round is a failure. A plain diagram with the arrows right is a success. So the prompt spends its budget on constraining content, and the little it spends on appearance goes to legibility rather than to beauty — "easy-to-read text", "avoid tiny text".
The explicit label list is the part most worth stealing. Eight molecules by name is a lot of prompt real estate, and it buys two things. It fixes the scope, so the model is not deciding whether to include FADH2. And it gives you a checklist: after generating, you read the eight names off the picture and confirm each one is present, spelled correctly, and attached to the right place. Without the list you would be checking the diagram against your own memory of the pathway, which is a much weaker audit.
This is also the template where the quality tier is a content decision rather than a budget one. Everything above it in the library runs at Standard. This one runs at HD because a diagram is mostly small text, and small text is precisely what a lower tier stops rendering cleanly. If you are producing teaching material, that is not an upgrade, it is the requirement.
Make a classroom diagram with the labels you specify — common questions
- Why HD for this one when the others are Standard?
- Small text. A diagram is typography with pictures attached, and label legibility is the first thing a lower quality tier gives up. OpenAI recommends `high` for dense labels and anything going into slides or course material.
- Do I still have to check the science?
- Yes, every time. Arrow direction and label placement are the common errors and both look convincing. Read your own label list off the finished image.
- Why landscape?
- Three sequential stages need horizontal room. In portrait the model stacks them and the connecting arrows start crossing each other.
- Can I match my school or brand style?
- Partly, and better in two passes: get the diagram correct first, then run an edit for the styling. Asking for both at once costs you accuracy.
Written and maintained by Andy SwiftPublished Last updated


