GPT Image

GPT Image models are useful when prompt adherence, text rendering, clean composition, or large output sizes matter. Pixapi exposes them through OpenAI-compatible image endpoints.

API surface

Operation	Endpoint	Mode
Image generation	POST /v1/images/generations	Sync
Image generation task	POST /v1/async/images/generations	Async
Image editing	POST /v1/images/edits	Sync
Image editing task	POST /v1/async/images/edits	Async

Model ids

Model id	Best for
gpt-image-2.5-flare	Speed-oriented generation, product variations, and high-volume creation.
gpt-image-2.5-sunburst	Quality-oriented compositions, campaign assets, and precise image editing.
gpt-image-2	OpenAI quality tiers, custom sizes, larger outputs, and image editing.
gpt-image-1.5	Fast GPT image generation and editing with strong text rendering.

Flare and Sunburst use the same request parameters, synchronous/asynchronous generation and editing endpoints, and credit pricing as GPT Image 2. Change only the model value to switch models.

Request fields

Send the common fields at the top level. Put the model-specific controls listed below inside providerOptions.

Field	Type	Required	Description
model	string	Yes	gpt-image-2, gpt-image-1.5, gpt-image-2.5-flare, gpt-image-2.5-sunburst. Case-sensitive model ID.
prompt	string	Yes	Describe the subject, composition, style, and requested edits. Pixapi keeps the first 2,000 characters; provide a non-empty prompt for both generation and editing.
n	integer	No	Default 1; only integer 1 is supported.
size	string	No	Use pixel dimensions WIDTHxHEIGHT, for example 1024x1024. Default auto. Custom dimensions: multiples of 16, longest edge ≤ 3840, aspect ratio ≤ 3:1, total pixels 655,360–8,294,400.
image	string / string[]	For edits	Public HTTP(S) image URL or URL array; required on /images/edits. Up to 14 URLs; Base64 input is not supported.
quality	string	No	auto (default), low, medium, or high. Higher quality can take longer.
providerOptions	object	No	Model-specific options; see the table below.

providerOptions

All fields in this table are nested inside providerOptions, not alongside model.

Field	Type	Description
output_format	string	Output image format: png (default, keeps transparency), jpeg (smaller files, no alpha), or webp (supports transparency).
background	string	Background mode: auto (default), opaque, or transparent. transparent can only be combined with output_format: "png" or output_format: "webp"; JPEG does not support an alpha channel.
moderation	string	Content moderation strictness: auto or low (more permissive). Default: auto for 1.5/2, low for 2.5.
output_compression	integer	Output compression level 0–100; applies only to jpeg and webp output.
mask_url	string	Public PNG mask URL for inpainting. Requires image; dimensions must match the first image. Transparent pixels mark the area to edit. Use a PNG with an alpha channel, under 4 MB.
nsfw_check	boolean	Optional prompt/image moderation; default false.

Input rules

Pixapi accepts one output and up to 14 reference images. Output pixels are selected with size.

Custom sizes must use the WIDTHxHEIGHT format (for example 1536x864) and satisfy all of the following limits at once:

Limit	Requirement
Multiple of 16	Both width and height must be divisible by 16.
Aspect ratio	Between 1:3 and 3:1.
Longest edge	No edge may exceed 3840px.
Minimum pixels	At least 655,360 (about 1024x640).
Maximum pixels	At most 8,294,400 (4K, equal to 3840x2160).

Because the ceiling is a total pixel count, 4K does not mean a 3840px long edge: on a square canvas the edge maxes out at 2880x2880 (√8,294,400 = 2880). Largest usable size per layout:

Layout	Size	Total pixels
Landscape	3840x2160	8,294,400
Portrait	2160x3840	8,294,400
Square	2880x2880	8,294,400
Ultra-wide 3:1	3840x1280	4,915,200

Outputs above 2560x1440 are experimental and their stability is not guaranteed.

Pricing

Model	Credits per image
gpt-image-2.5-flare	from 1
gpt-image-2.5-sunburst	from 1
gpt-image-2	from 1
gpt-image-1.5	from 1

gpt-image-1.5 is billed for 1K sizes only (1024x1024, 1536x1024, 1024x1536, or auto): low/medium/high is 1/4/14. It does not support 2K or 4K outputs.

For gpt-image-2, gpt-image-2.5-flare, and gpt-image-2.5-sunburst, credits follow the size and quality matrix below. quality=auto is billed as high, and size=auto defaults to the 1K tier.

Size tier	Long edge	low	medium	high / auto
1K	<= 1536px	1	4	14
2K	<= 2048px	2	17	67
4K	<= 3840px	4	34	133

The displayed costCredits for a request equals the matrix value times n.

Generate example

const response = await fetch('https://api.pixapi.ai/v1/images/generations', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${process.env.PIXAPI_KEY}`,
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'gpt-image-2.5-flare',
    prompt: 'A clean SaaS hero image showing API model orchestration',
    n: 1,
    size: '1536x1024',
    quality: 'high',
  }),
});

const image = await response.json();

Edit example

curl https://api.pixapi.ai/v1/images/edits \
  -H "Authorization: Bearer $PIXAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.5-sunburst",
    "prompt": "Replace the background with a bright studio setup",
    "image": "https://cdn.example.com/input.png",
    "size": "auto",
    "quality": "auto"
  }'

Response

{
  "created": 1766880000,
  "data": [
    {
      "url": "https://cdn.pixapi.ai/generated/gpt-image.png"
    }
  ]
}