Can I trust the technical labels?

No — check them. OpenAI's own guidance for this technique says to verify labels and factual relationships, not just appearance. Treat the output as a well-drawn first draft.

Why is the prompt so short?

Because the model already knows what is inside a coffee machine. The prompt only has to decide which parts to show and who the picture is for.

What quality should I use?

Standard is what the official example used. Move to HD as soon as the diagram carries more than a few labels — small text degrades before anything else does.

How do I get it in another language?

Generate it in English first, then run the translate-and-keep-layout edit on the result. Asking for a second language in the first pass tends to cost you layout quality as well.

GPT Image 2.5 prompt — an infographic that explains how something works

Four lines of prompt, and the model supplies the engineering it already knows.

A labelled cutaway infographic of an automatic coffee machine showing bean hopper, grinder, water tank and boiler

Settings the Turn a process you half-understand into a diagram prompt ran at
Mode	Model	Quality	Size	Ratio	Credits
Text to Image	Think	Standard	1K	2:3	6 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 1024x1536, a legal size we do not currently offer; the frame here is the same shape at our nearest size.

A cutaway of an automatic coffee machine, from bean hopper to grinder to boiler. The prompt is remarkably short because it is not describing a picture — it is naming a real machine and asking for the flow. What the model brings is the domain knowledge; what you bring is the decision about which parts matter and who is looking at it.

In full

The prompt, nothing cut

Create a detailed Infographic of the functioning and flow of an automatic coffee machine like a Jura.
From bean basket, to grinding, to scale, water tank, boiler, etc.
I'd like to understand technically and visually the flow.

225 characters

Run it here

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 · Standard · 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.

3 levers

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 reference
"like a Jura" is four characters of context that pins down the whole form factor. Any prompt about a real category of object gets sharper when you name a specific instance of it — not to copy that product, but to fix the model on one interpretation instead of averaging several.
The waypoint list
"From bean basket, to grinding, to scale, water tank, boiler, etc." is the running order. Replace it with your own stages and the layout follows. The "etc." matters too — it grants permission to add the steps you forgot.
The stated purpose
"I'd like to understand technically and visually the flow" tells the model what the picture is for. Swap it for "for a service manual" or "for a customer leaflet" and the density, the labelling and the level of jargon all move.

Three ways to break it

Each one is a real failure with a reason attached. A rule without a reason is not usable.

01Trusting the labels without reading them
This is the one that costs people real embarrassment. Diagrams are the place where a plausible-looking image can be factually wrong, and OpenAI says so directly: verify labels and factual relationships as well as appearance. The picture is a draft of a diagram, not a diagram.
02Running it at Draft or Standard when the labels are dense
Small text is the first casualty of a lower quality tier. If your diagram has more than a handful of callouts, HD is the floor — the official guide moves to `high` for exactly this reason on its denser examples.
03Asking for the diagram and the branding in one pass
A prompt that also specifies a logo, a colour system and a footer is spending the model's attention on layout furniture instead of on the flow. Get the flow right first, then edit the styling in a second turn.

Why it is written this way

Inside the Turn a process you half-understand into a diagram prompt

The interesting thing about this prompt is how little of it is visual. There is no palette, no style word, no composition instruction, no aspect guidance beyond the size parameter. It names a machine, lists the stages, and states what the reader wants to understand. That is closer to a brief you would hand an illustrator than to what most people write when prompting an image model.

It works because a technical diagram is a knowledge problem before it is a drawing problem. The hard part of drawing a coffee machine cutaway is knowing that there is a grinder, that it feeds a tamper, that water goes through a boiler at a particular point in the sequence. The model has that knowledge. What it does not have is your judgement about which subset of it belongs in this picture, which is exactly what the waypoint list supplies.

The phrase "like a Jura" is worth isolating. Naming a specific product is not about reproducing that product — it is a disambiguation device. "Automatic coffee machine" spans everything from a pod machine to a commercial espresso rig, and a model asked to draw the average of those will draw something that is none of them. One brand name collapses that distribution to a single form factor, and the rest of the prompt then has something concrete to attach to.

The caveat has to be stated plainly because it is the failure mode that matters. An infographic that looks right and is wrong is worse than one that looks obviously unfinished, because it will get used. Arrows can point the wrong way. A component can be labelled as its neighbour. Water can appear to flow uphill through a boiler. None of that is visible unless you know the subject, which means this template is for people who can check the output, not for people who need the output because they cannot.

For anything going into teaching material, slides or a manual, treat the result as a layout you now have to fact-check line by line — and budget for at least one editing turn to fix the labels the first pass got wrong.

4 questions

Turn a process you half-understand into a diagram — common questions

Can I trust the technical labels?: No — check them. OpenAI's own guidance for this technique says to verify labels and factual relationships, not just appearance. Treat the output as a well-drawn first draft.
Why is the prompt so short?: Because the model already knows what is inside a coffee machine. The prompt only has to decide which parts to show and who the picture is for.
What quality should I use?: Standard is what the official example used. Move to HD as soon as the diagram carries more than a few labels — small text degrades before anything else does.
How do I get it in another language?: Generate it in English first, then run the translate-and-keep-layout edit on the result. Asking for a second language in the first pass tends to cost you layout quality as well.