GPT Image 2.5 prompt — translate an image and keep the design identical
One sentence. The second half of it is what stops the design being redrawn.
| Mode | Model | Quality | Size | Ratio | Credits |
|---|---|---|---|---|---|
| Image to Image | Think | HD | 1K | 2:3 | 28 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.
The coffee-machine infographic, in Spanish. The prompt is fourteen words long and half of it is a prohibition, which is the correct ratio for this job — the model already knows how to translate, and what you are actually buying is restraint. This is also the cheapest way to localise a graphic that exists only as a flat image.
The prompt, nothing cut
Translate the text in the infographic to Spanish. Do not change any other aspect of the image.
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 target language
Swap Spanish for anything. Check the result harder for scripts the source layout was not designed for — a language that runs longer than English will either shrink the type or overflow the box it was given.
The scope word
"the text in the infographic" scopes the change. Narrow it further — "only the callout labels", "only the title" — when part of the copy has to survive untranslated.
The quality tier
This one runs at HD rather than Standard, and it is the same reason as the original diagram: the whole point of the output is small legible type in a language you may not read.
Three ways to break it
Each one is a real failure with a reason attached. A rule without a reason is not usable.
01Dropping "Do not change any other aspect of the image"
Without it the model treats the request as "make a Spanish version of this infographic" and redraws it. You get a Spanish infographic, but not yours — different icons, different arrangement, different everything.
02Not reading the output
Some words survive untranslated, and they are usually the ones inside graphical elements rather than in text layers. OpenAI's own note on this technique says to check the translation and any words left in the original language.
03Running it on a photograph of text
This works on a graphic whose text is a designed layer. On a photograph — a street sign, a page in a book — you are asking for a much harder composite and the surrounding pixels will move.
Inside the Translate the text in an image without moving the layout prompt
Fourteen words, and it is one of the most commercially useful prompts in the library. Anyone who has ever needed a marketing graphic in a second language knows the real cost is not the translation — it is that the graphic lives as a flat PNG somewhere and the person who had the layered file left two years ago.
The structure is worth naming because it recurs across every editing template here: one clause states the change, one clause states the scope of what must not change. "Translate the text in the infographic" and "Do not change any other aspect of the image". OpenAI's prompting fundamentals put this as separating changes from constraints, and in image editing the constraint half is consistently the harder one to remember and the more important one to include.
The reason it matters so much here is a subtle one about how the request is read. "Translate this infographic to Spanish" is a perfectly reasonable instruction that a competent designer would fulfil by producing a Spanish infographic — possibly a better one, possibly rearranged to suit the longer Spanish strings. That is not wrong. It just is not what you asked for, and you will not notice how much moved until you put the two side by side. The prohibition converts an open brief into a constrained edit.
The failure that survives the prohibition is partial translation. Words baked into a graphical element — a label that is part of an icon, text on a rendered surface — behave differently from words in a text layer, and some of them come back in English. This is not a prompt problem you can fix by rewording; it is something to check for. Read every string in the output, including the ones inside pictures.
And run it at HD. The entire value of the artefact is legible type, quite possibly in a language you cannot proofread by glancing.
Translate the text in an image without moving the layout — common questions
- Why is the prohibition necessary?
- Because "translate this infographic" is a reasonable brief that a good designer would answer by redrawing it. The prohibition turns an open brief into a constrained edit.
- Will every word get translated?
- Not reliably. Text baked into graphical elements often survives in the original language. Read every string in the output, including the ones inside icons.
- Does it work on photographs?
- Much less well. This is built for graphics whose text sits in a designed layer. A photograph of text is a composite job and the surrounding pixels will move.
- What about languages that run longer than English?
- Expect the type to shrink or the box to overflow, because the layout was designed around the original string lengths. Check the tightest text box first.
Other prompts that start the same way
- Think · Standard · 1K · 2:3
Input 
Result Borrow one image's style and draw something else with it
Assigns the reference a single role — style — then describes the new subject separately.
Bring an image10 cr - Think · Standard · 1K · 2:3
Input 
Result Change what someone is wearing without changing their face
Nine preservation clauses and nine prohibitions, because each one closes a known drift.
Bring an image10 cr - Think · Standard · 1K · 2:3
Input 
Result Put something from one photo into another
Numbers both inputs, names the element and its destination, then forbids everything else.
Bring an image10 cr
Written and maintained by Andy SwiftPublished Last updated