Skip to main content

Filtered — This model refuses requests beyond its own limits outright. Measured, not assumed — models we have not measured carry no label at all.Browse every model at this filtering level
openai/gpt-image-2/text-to-image

OpenAI's GPT Image 2 text-to-image, which can write legible text into the picture. Choose low, medium or high quality, a 1K, 2K or 4K size tier and one of 15 aspect ratios, and get 1 to 10 images as PNG, JPEG or WebP.

From $0.0018/image20% below the publisher's list price
Run model

Workbench

Input

Text to image
377/5000
Up to 5,000 characters. Put every word that must appear in the image in double quotes and say where it goes; the rest is read as description. The first 100 input tokens of the prompt are included in the image price; longer prompts add a small charge per extra token, shown in the quote. Chinese, Japanese and Korean text counts about one token per character, other text about one token per four characters. at least 1 characters · at most 5000 characters
affects price
Fifteen shapes from 1:3 to 3:1, 1:1 by default. The shape changes the price: a wide or tall frame costs less than a square at the same tier.
affects price
1k (default), 2k or 4k. A square frame is 1024, 2048 or 2880 pixels across; a larger tier costs more per image. On the economy route a 1k image can come out somewhat larger.
Qualityaffects price
lowmediumhigh
low, medium (default) or high. OpenAI bills by output tokens, so each step up costs about four to nine times the one below at the same size: low for drafts, medium for most finals, high when fine type and skin must hold up close.
Output format
PNGJPEGWebP
png (default), jpeg or webp. Does not change the price.
Output quality
Compression quality for jpeg and webp, 0 to 100: higher keeps more detail and makes larger files. Not used for png, and it does not change the price. no lower than 0 · no higher than 100
Official fallback
On by default. A request for a single medium-quality 1k PNG in any shape except 21:9, 9:21, 3:1 and 1:3 runs on the economy route first; if that route cannot complete it, for example because it refuses the prompt, the request runs again on the official route and is billed at the official price instead of the economy price. Every other request runs on the official route from the start. Turn it off to pay only the economy price: requests the economy route cannot serve are then rejected, and one it takes may still be refused or fail.
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
neon-posterExample result
Animate
Edit
Split into layers

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

OPENAI IMAGE API

GPT Image 2 API at a glance

OpenAI's GPT Image 2 with every generation and editing control; medium 1K images start well below OpenAI's list price.

Economy route
1K
Medium-quality 1K images in eleven shapes start on a lower-priced route; Official fallback reruns anything it cannot finish on the official route.
Largest size tier
4K
A 4k square renders at 2880 × 2880; a 16:9 frame at 3840 × 2160.
Images per request
10
Ask for up to ten variations of one prompt in a single call.
Edit sources
10
Combine up to ten source images, plus an optional mask for inpainting.

Results

What you can make with GPT Image 2

  • A film poster: a woman in a cropped red leather jacket against a rain-streaked phone booth under pink and teal neon, the title AFTER HOURS across the top

    Typography

    Titles that come back spelled right

    A late-night film poster with a condensed title across the top and a spaced line of small capitals at the bottom, generated in one call at medium quality. Both lines came back exactly as quoted in the prompt.

    Text to image, medium quality, 2:3.

  • The same cover retitled SALT & LINEN with the lines NEW SEASON and THE ISLAND EDIT, the man now in a navy linen shirt

    Editing

    Change the words, keep the photograph

    The magazine cover from the text-to-image example, retyped: a new masthead, two new cover lines and a navy shirt, while his face, the boat and the golden light stayed as they were.

    Edit, one source, input fidelity high.

  • A film poster titled TWO TIDES: a woman in a red leather jacket and a man in a white shirt and dark coat standing back to back on a wet neon pier

    Multi-image

    Two sources, one new frame

    The woman from the poster and the man from the cover, placed back to back on a neon pier under a new title. Each source is named in the prompt by its order.

    Edit, two sources, 2:3.

Content test

Is GPT Image 2 uncensored? What we tested

We ran each level on our live setup and checked every clip frame by frame. A pass means the content was actually on screen, not just that the request went through.

Content levels tested on GPT Image 2, by mode
LevelImage editingText to image
MildThe lightest prompts in the test
RefusedTested Sep 29, 2026
RefusedTested Sep 29, 2026
ModerateWhere filtered models start to refuse
Not tested yet
Not tested yet
StrongThe middle of the scale
RefusedTested Sep 29, 2026
RefusedTested Sep 29, 2026
Very strongWhere most models start to hold back
Not tested yet
Not tested yet
MaximumThe top of the scale
Not tested yet
Not tested yet

Renders: on screen as written. Softened: generated, but milder than what was written. Refused: rejected. Not tested yet: we have not run that level on that mode.

How each mode behaves

All modes
OpenAI's safety system screens every prompt and every result. In our tests every content level we sent was refused, Mild included, even at the most permissive moderation level OpenAI offers; fashion, evening wear and portraits went through. Refused requests are not charged.
Image editing
GPT Image 2 cannot return a transparent background; OpenAI offers that on GPT Image 2.5. With aspect ratio auto the output follows the shape of the first source at about 1.5 megapixels and the size tier is not applied.

Use cases

What to build with the GPT Image 2 API

  • Posters and covers with real type

    Film posters, magazine covers and event flyers where the title, the date and the small capitals have to read correctly without a designer retyping them.

  • Campaign variations

    Up to ten images per request from one brief; then edit the chosen one with a mask instead of re-rolling the whole frame.

  • Retouching and restyling

    Change a garment, a colour or a line of type in an existing photograph while the face, the pose and the light stay put.

Prompts

GPT Image 2 prompt examples you can copy

Paste one into the playground as it is, then change one thing at a time.

  • Text to image

    Titles that come back spelled right

    A late-night film poster. A woman in a cropped red leather jacket and a black slip top leans against a rain-streaked phone booth under pink and teal neon, looking straight into the lens, hair damp, city bokeh behind her. Across the top, a bold condensed title reads "AFTER HOURS"; along the bottom in small spaced capitals: "ONE NIGHT ONLY · DOORS AT 11". Cinematic 35mm grain.

    Why it works. Both lines of type are quoted and given a position. Medium quality at the 1k tier was enough for the title and the small capitals to come back spelled correctly.

    Try this prompt

  • Image editing

    Change the words, keep the photograph

    Retype this magazine cover. Change the masthead to read "SALT & LINEN" in the same tall serif, and change the two cover lines to "NEW SEASON" and "THE ISLAND EDIT" in the same small bold sans-serif. Change his white linen shirt to a washed navy linen shirt. Keep his face, pose, the boat, the sea and the golden light exactly as they are.

    Why it works. The prompt lists what changes (masthead, two cover lines, shirt colour) and then everything that must stay: face, pose, boat, sea and light.

    This mode starts from your own image: add it in the playground before you run.

    Try this prompt

  • Image editing

    Two sources, one new frame

    Make a new film poster from these two images: the woman from Image 1 and the man from Image 2 stand back to back on a wet neon-lit pier at night, both looking at the camera, she keeps her red leather jacket and he wears his white linen shirt under a dark coat. Title across the top in a bold condensed face: "TWO TIDES". Bottom line in small spaced capitals: "IN CINEMAS THIS WINTER".

    Why it works. Two sources, called the first and the second image. Each person keeps the outfit from their own source, and the new title and bottom line are quoted.

    This mode starts from your own image: add it in the playground before you run.

    Try this prompt

Pricing

GPT Image 2 API pricing

One image starts at $0.0018 at 1K and $0.0037 at 4K.

The playground shows the exact price for your settings before you run. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.

See every price tier

Get started

How to use the GPT Image 2 API in three steps

  1. Step 1

    Quote the words you need

    Put every headline, label or cover line in double quotes and say where it sits. In our poster and cover tests every quoted line came back spelled correctly at medium quality.

  2. Step 2

    Draft low, finish medium or high

    Low costs a small fraction of medium at the same size and is enough to judge the composition. Move to medium or high for the final, where small type and skin texture sharpen.

  3. Step 3

    Pick the shape before the tier

    A 2:3 or 16:9 frame costs less than a square at the same tier, so choose the ratio you will publish in first, then decide between 1k, 2k and 4k.

Open the playgroundGet an API key

Parameter reference and code samples are in the API and Readme tabs at the top of this page.

Why SpicyAPI

Why run GPT Image 2 on SpicyAPI

  • No filter from us

    We add no content filter of our own. What the model renders is what you get back.

  • Tested, not guessed

    Content labels come from real generations we ran and checked, with the test date on the page.

  • Pay per result, in dollars

    No subscription and no credits. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.

  • One key for every model

    One API key and balance cover image, video, audio and chat models, so you can explore more models from the same account.

FAQ

GPT Image 2 API FAQ

About GPT Image 2

GPT Image 2 is OpenAI's image model: legible text in images, editing with up to ten sources and a mask, fifteen aspect ratios from 1K to 4K and up to ten images per request, with every setting priced below OpenAI's list price. OpenAI's content policy applies, so it is a tool for design, advertising and editorial work rather than anything its filter would stop.

All GPT Image models3

Related models and next tasks