Introduction
GPT Image 2 API creates posters and magazine covers with readable headlines, then revises their wording or combines up to 10 references. This April 2026 model offers low, medium and high quality, fifteen aspect ratios and masked editing, without transparent output. GPT Image 2 accepts a PNG mask sized to the first reference, with transparent regions marking changes and opaque regions marking areas to retain; without a mask, the prompt specifies the scope.
| Field | Value |
|---|---|
| Input | Image |
| Output | Image |
| Model | OpenAI gpt-image-2 (snapshot 2026-04-21) |
| Endpoints | GPT Image 2 — Text to image; image editing with 1 to 10 sources and an optional mask |
| Quality tiers | GPT Image 2 — low, medium (default), high |
| Output size | GPT Image 2 — 1k, 2k or 4k tier; 1:1 at 1024, 2048 or 2880 pixels square |
| Aspect ratios | GPT Image 2 — 15 shapes from 1:3 to 3:1 |
| Images per request | GPT Image 2 — 1 to 10 |
| Output formats | GPT Image 2 — PNG (default), JPEG, WebP; compression 0 to 100 for JPEG and WebP |
| Content policy | OpenAI's content policy; even Mild-level requests are refused |
Key features & highlights
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.
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.
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.
Request parameters
| Parameter | Required | Description |
|---|---|---|
prompt | Yes | 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. at least 1 characters · at most 5000 characters |
image_urls | Yes | Upload 1–10 source images for GPT Image 2 editing and identify them by position; under standard pricing each output includes a fixed input fee for each reference. at least 1 items · at most 10 items |
mask_url | No | GPT Image 2 accepts a PNG mask sized to the first reference, with transparent regions marking changes and opaque regions marking areas to retain; without a mask, the prompt spec… |
aspect_ratio | No | GPT Image 2 generation has 15 ratios from 1:3 to 3:1, initially 1:1; edits also offer auto, which follows the first image at about 1.5 megapixels and ignores resolution, while f… |
resolution | No | GPT Image 2 defaults to 1k and also offers 2k or 4k; squares are 1024, 2048 or 2880 pixels wide, with higher tiers priced higher and some lower-priced 1k results slightly larger. |
quality | No | GPT Image 2 has low, medium and high quality, with medium selected initially; output-token costs rise roughly four to ninefold per step at a fixed size. |
num_outputs | No | Set the GPT Image 2 batch size to 1–10, initially 1; each poster or cover returned counts as a billable image. |
input_fidelity | No | For GPT Image 2 cover revisions, high input fidelity helps preserve faces, fabric and lettering; the default is low, and choosing high does not change the price. |
background | No | GPT Image 2 edits accept auto or opaque backgrounds, initially auto; transparent delivery is unavailable here, while GPT Image 2.5 offers cut-outs. |
output_format | No | GPT Image 2 can deliver png, jpeg or webp at the same rendering price, with png selected initially. |
output_quality | No | Set JPEG/WebP compression quality between 0 and 100 for GPT Image 2 exports; higher values keep more detail in larger files, while PNG and the image price are unaffected. |
official_fallback | No | GPT Image 2 is billed per output by quality, resolution and shape; standard-priced edits add a fixed source-image charge to each output, and wide or tall shapes cost less than s… Default: |
Pricing
GPT Image 2 is billed per output by quality, resolution and shape; standard-priced edits add a fixed source-image charge to each output, and wide or tall shapes cost less than squares at the same tier. The lower-priced GPT Image 2 offer covers one medium 1k PNG, in any ratio except 21:9, 9:21, 3:1 and 1:3, and for edits only with no mask, input_fidelity low and background auto. With Official fallback enabled by default, a lower-priced request that is refused or cannot finish is retried at the standard price; the GPT Image 2 quote shows the starting amount and maximum, other settings use standard pricing immediately, and disabling fallback rejects settings outside the offer without guaranteeing success for eligible ones. A refused GPT Image 2 request is not charged; under standard pricing only, prompt tokens after the first 100 add a fee once per request, estimated at one token per Chinese, Japanese or Korean character and about one per four other characters.
| Specification | Price |
|---|---|
| $0.001816 / image |
| $0.001816 / image |
| $0.00244 / image |
| $0.00244 / image |
| $0.002752 / image |
| $0.002752 / image |
| $0.002872 / image |
| $0.002872 / image |
| $0.00304 / image |
| $0.00304 / image |
| $0.003568 / image |
| $0.003568 / image |
| $0.003736 / image |
| $0.003736 / image |
| $0.003856 / image |
| $0.003856 / image |
| $0.004192 / image |
| $0.004192 / image |
| $0.004456 / image |
| $0.004456 / image |
| $0.005104 / image |
| $0.005152 / image |
| $0.005152 / image |
| $0.005704 / image |
| $0.005704 / image |
| $0.005896 / image |
| $0.006016 / image |
| $0.006016 / image |
| $0.006712 / image |
| $0.006712 / image |
| $0.007312 / image |
| $0.007312 / image |
| $0.0076 / image |
| $0.0076 / image |
| $0.008128 / image |
| $0.008128 / image |
| $0.009304 / image |
| $0.009304 / image |
| $0.009928 / image |
| $0.009928 / image |
| $0.011176 / image |
| $0.011176 / image |
| $0.01192 / image |
| $0.01192 / image |
| $0.01324 / image |
| $0.01324 / image |
| $0.016216 / image |
| $0.016216 / image |
| $0.01396 / image |
| $0.01396 / image |
| $0.018592 / image |
| $0.018592 / image |
| $0.020368 / image (With Official fallback: up to $0.021472) |
| $0.020368 / image (With Official fallback: up to $0.021472) |
| $0.020998 / image (With Official fallback: up to $0.023992) |
| $0.020998 / image (With Official fallback: up to $0.023992) |
| $0.022 / image (With Official fallback: up to $0.028) |
| $0.022 / image (With Official fallback: up to $0.028) |
| $0.022864 / image (With Official fallback: up to $0.031456) |
| $0.022864 / image (With Official fallback: up to $0.031456) |
| $0.023626 / image (With Official fallback: up to $0.034504) |
| $0.023626 / image (With Official fallback: up to $0.034504) |
| $0.024112 / image |
| $0.024112 / image |
| $0.025636 / image (With Official fallback: up to $0.042544) |
| $0.027448 / image (With Official fallback: up to $0.049792) |
| $0.032272 / image |
| $0.032272 / image |
| $0.03688 / image |
| $0.03688 / image |
| $0.043072 / image |
| $0.043072 / image |
| $0.048064 / image |
| $0.048064 / image |
| $0.050776 / image |
| $0.050776 / image |
| $0.055432 / image |
| $0.055432 / image |
| $0.062416 / image |
| $0.062416 / image |
| $0.0652 / image |
| $0.0652 / image |
| $0.067984 / image |
| $0.067984 / image |
| $0.080464 / image |
| $0.080464 / image |
| $0.086032 / image |
| $0.086032 / image |
| $0.094288 / image |
| $0.094288 / image |
| $0.103984 / image |
| $0.103984 / image |
| $0.11296 / image |
| $0.11296 / image |
| $0.14272 / image |
| $0.14272 / image |
| $0.054616 / image |
| $0.054616 / image |
| $0.07144 / image |
| $0.07144 / image |
| $0.082936 / image |
| $0.082936 / image |
| $0.095248 / image |
| $0.095248 / image |
| $0.096472 / image |
| $0.096472 / image |
| $0.110728 / image |
| $0.110728 / image |
| $0.124624 / image |
| $0.124624 / image |
| $0.127864 / image |
| $0.127864 / image |
| $0.135064 / image |
| $0.135064 / image |
| $0.142816 / image |
| $0.142816 / image |
| $0.168976 / image |
| $0.171016 / image |
| $0.171016 / image |
| $0.19456 / image |
| $0.19456 / image |
| $0.197104 / image |
| $0.197104 / image |
| $0.197968 / image |
| $0.220456 / image |
| $0.220456 / image |
| $0.248464 / image |
| $0.248464 / image |
| $0.259552 / image |
| $0.259552 / image |
| $0.274216 / image |
| $0.274216 / image |
| $0.320608 / image |
| $0.320608 / image |
| $0.342928 / image |
| $0.342928 / image |
| $0.37588 / image |
| $0.37588 / image |
| $0.414736 / image |
| $0.414736 / image |
| $0.456544 / image |
| $0.456544 / image |
| $0.569656 / image |
| $0.569656 / image |
Use cases
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.
Prompt guide & pro tips
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.
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.
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.
GPT Image 2
GPT Image 2 API creates posters and magazine covers with readable headlines, then revises their wording or combines up to 10 references. This April 2026 model offers low, medium and high quality, fifteen aspect ratios and masked editing, without transparent output.
| a written brief | GPT Image 2 Text to image | | an image to change | GPT Image 2 Edit |
Workflow
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.
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.
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.
Controls
GPT Image 2 has low, medium and high quality, with medium selected initially; output-token costs rise roughly four to ninefold per step at a fixed size.
GPT Image 2 defaults to 1k and also offers 2k or 4k; squares are 1024, 2048 or 2880 pixels wide, with higher tiers priced higher and some lower-priced 1k results slightly larger.
Upload 1–10 source images for GPT Image 2 editing and identify them by position; under standard pricing each output includes a fixed input fee for each reference.
GPT Image 2 accepts a PNG mask sized to the first reference, with transparent regions marking changes and opaque regions marking areas to retain; without a mask, the prompt specifies the scope. Review the lettering around a GPT Image 2 masked edit; the mask does not guarantee that every boundary is followed exactly.
GPT Image 2 edits accept auto or opaque backgrounds, initially auto; transparent delivery is unavailable here, while GPT Image 2.5 offers cut-outs.
For GPT Image 2 cover revisions, high input fidelity helps preserve faces, fabric and lettering; the default is low, and choosing high does not change the price.
Examples
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.
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.
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.
GPT Image 2 does not provide NSFW generation: SpicyAPI’s recorded tests rejected even Mild at OpenAI’s most permissive moderation setting, although fashion and portraits succeeded. OpenAI reviews both prompts and outputs; rejected GPT Image 2 jobs are uncharged, SpicyAPI adds no filtering and cannot override these checks, and the catalogue identifies unrestricted alternatives.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "openai/gpt-image-2/edit",
"input": {
"aspect_ratio": "auto",
"background": "auto",
"image_urls": [
"https://example.com/reference.png"
],
"input_fidelity": "low",
"num_outputs": 1,
"output_format": "png",
"prompt": "A ceramic teapot on a linen cloth, soft window light, muted palette"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.







