How do I stop the model inventing extra text?

Say so explicitly. "No extra text, no watermarks, no unrelated logos" at the end of the prompt is the shortest reliable form.

What if my brand name is unusual?

Spell it letter by letter in the prompt. OpenAI recommends this directly, and it is the difference between four consistent renders and four spellings.

Can I ask for a specific font?

You can ask, but you cannot rely on it. Describe weight, contrast and placement instead — those transfer, and a font name mostly does not.

Standard or HD for text?

The official example is Standard, and a short tagline at a large size survives it. Compare Standard against HD as soon as the type gets small, dense, or uses more than one face.

GPT Image 2.5 prompt — an ad with the tagline rendered exactly once

Quote the copy, say how many times it appears, and forbid everything else.

A streetwear campaign photograph of friends together, carrying the tagline "Yours to Create."

Settings the Put an exact tagline in an ad and have it spelled right 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 streetwear campaign image for a made-up brand called Thread, carrying the line "Yours to Create." Text rendering is the capability GPT Image 2.5 is sold on, and this prompt shows the three habits that make it reliable: put the wording in quotes, state the number of times it should appear, and explicitly ban the extra text the model would otherwise invent to fill a layout.

In full

The prompt, nothing cut

Give me a cool in culture ad / fashion shot for a brand called Thread.
It's a hip young street brand. The ad shows a group of friends hanging out together with the tagline "Yours to Create."
Make it feel like a polished campaign image for a youth streetwear audience: stylish, contemporary, energetic, and tasteful.
Use clean composition, strong color direction, natural poses, and premium fashion photography cues.
Render the tagline exactly once, clearly and legibly, integrated into the ad layout.
No extra text, no watermarks, no unrelated logos.

550 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 quoted string
Anything inside quotes is treated as copy to reproduce rather than a description to interpret. This is the difference between getting "Yours to Create." and getting a plausible-looking tagline that means roughly the same thing.
The repeat count
"exactly once" is not padding. Left unstated, a model laying out an ad will often repeat the line — small in a corner, large across the middle — because that is what a lot of advertising does. Say the number.
The audience line
"for a youth streetwear audience" is doing more for the look than any of the adjectives around it. Swap the audience and the wardrobe, the location, the colour direction and the casting all move together, which is far more efficient than specifying each one.

Three ways to break it

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

01Leaving the exclusion list off
"No extra text, no watermarks, no unrelated logos" is the line that keeps invented brand marks and stray sub-headlines out of the frame. Advertising imagery in the training data is full of both, so the model will supply them unless told not to.
02Asking for a long tagline
Short strings render reliably; sentences do not. Every extra word is another chance for a malformed glyph, and the failure is not graceful — one wrong letter makes the whole image unusable for its purpose.
03Spelling an unusual brand name normally
For invented or unusual names, OpenAI recommends spelling it letter by letter in the prompt. "Thread" is a common word and survives; something like "Kyrbo" will come back as three different spellings across four generations unless you spell it out.

Why it is written this way

Inside the Put an exact tagline in an ad and have it spelled right prompt

Text rendering is the headline capability of this model generation, and it is also the one where the gap between "usually works" and "works" is commercially decisive. An ad with a typo in the tagline is not a slightly worse ad; it is a file you throw away. So the prompt is built around three habits, and all three are about removing the model's discretion.

The first is quoting. Text inside quotation marks is treated as a string to reproduce; text described in prose is treated as an idea to express. Write "with a tagline about creativity" and you get a tagline about creativity — a different one each time. Write "Yours to Create." and you get that.

The second is counting. This is the habit people skip, and it produces the most common failure in text-carrying layouts: the same line appearing twice at different sizes. It happens because it is genuinely common in real campaign imagery, so the model is not wrong to consider it — it just was not asked. "Render the tagline exactly once" removes the question.

The third is exclusion. The last line of the prompt does nothing positive at all; it only forbids. That is a good use of the last line. Ad imagery in any training corpus is dense with legal text, sub-headlines, sponsor logos and agency watermarks, and a model producing "a polished campaign image" will reach for those props because they are part of what makes the genre read as itself. Naming them as unwanted costs eight words and saves a rerun.

One thing this prompt does not do, and should not: it does not specify a typeface. Asking for a named font is asking for a thing the model cannot guarantee, and it burns attention that would otherwise go into getting the letters right. Describe the treatment — weight, contrast, placement — and let the shapes fall where they fall.

4 questions

Put an exact tagline in an ad and have it spelled right — common questions

How do I stop the model inventing extra text?: Say so explicitly. "No extra text, no watermarks, no unrelated logos" at the end of the prompt is the shortest reliable form.
What if my brand name is unusual?: Spell it letter by letter in the prompt. OpenAI recommends this directly, and it is the difference between four consistent renders and four spellings.
Can I ask for a specific font?: You can ask, but you cannot rely on it. Describe weight, contrast and placement instead — those transfer, and a font name mostly does not.
Standard or HD for text?: The official example is Standard, and a short tagline at a large size survives it. Compare Standard against HD as soon as the type gets small, dense, or uses more than one face.