Image

GPT Image 2 image

OpenAI GPT Image 2 — text-to-image with optional references. Quality low/medium/high; 1K/2K/4K. Returns one image per request.

Category: Create · gpt-image-2

How to use & limits

Text-to-image with optional references to guide generation.
prompt accepts up to 20,000 characters.
quality: low, medium or high.
resolution: 1K, 2K or 4K.
size: auto or aspect ratios (1:1, 2:3, 3:2, 4:3, 3:4, 4:5, 5:4, 16:9, 9:16, 21:9).
background auto/transparent/opaque; output_format jpeg/png/webp.
Returns one image per request.

Capabilities

What this model supports, straight from the schema — no pricing, no infrastructure detail.

Capability	Value
`n_max` Max variations per job	1
`create_image` Text-to-image	Yes
`allowed_quality` Quality levels	low, medium, high
`reference_images` Reference images	Yes
`max_prompt_length` Max prompt length	20000
`allowed_resolutions` Allowed resolutions	1K, 2K, 4K
`max_reference_images` Max reference images	16

Parameters

Fields accepted in params (or first-class, where applicable) for POST /generations.

Name	Label	Type	Required	Default	Options	Min / Max	Description
`prompt`	Prompt	textarea	Required	—	—	—	Image description, up to 20,000 characters. Required.
`quality`	Quality	select	Optional	low	low · medium · high	—	Output quality: low, medium or high. Default low.
`image_urls`	Optional references	multi_select	Optional	—	—	—	Optional references to guide generation.
`size`	Size	select	Optional	auto	auto · 1:1 · 2:3 · 3:2 · 4:3 · 3:4 · 4:5 · 5:4 · 16:9 · 9:16 · 21:9	—	auto or aspect ratio (1:1, 2:3, 3:2, 4:3, 3:4, 4:5, 5:4, 16:9, 9:16, 21:9). Default auto.
`resolution`	Resolution	select	Optional	1K	1K · 2K · 4K	—	Resolution: 1K, 2K or 4K. Default 1K. size=auto forces 1K.
`output_format`	Output format	select	Optional	png	jpeg · png · webp	—	File format: jpeg, png or webp. Default png.
`background`	Background	select	Optional	auto	auto · transparent · opaque	—	Background: auto, transparent or opaque. Default auto.

Example request

bash

curl -X POST https://caranguejo.art/api/v1/dev/generations \
  -H "Authorization: Bearer $CK" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "gpt-image-2",
  "prompt": "Premium product photo of a silver espresso machine on a white background.",
  "params": {
    "quality": "high",
    "size": "16:9",
    "resolution": "4K"
  }
}'

javascript

const res = await fetch("https://caranguejo.art/api/v1/dev/generations", {
  method: "POST",
  headers: {
    "Authorization": "Bearer " + process.env.CK,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    "model": "gpt-image-2",
    "prompt": "Premium product photo of a silver espresso machine on a white background.",
    "params": {
      "quality": "high",
      "size": "16:9",
      "resolution": "4K"
    }
  }),
});
const { data } = await res.json();

Example response

json

{
  "id": "9f0c1a7e-…",
  "object": "generation",
  "model": "gpt-image-2",
  "type": "image",
  "status": "completed",
  "output": {
    "assets": [
      {
        "url": "https://media.caranguejo.art/outputs/example.png",
        "type": "image"
      }
    ],
    "images": [
      {
        "url": "https://media.caranguejo.art/outputs/example.png"
      }
    ]
  }
}

The job's real cost is returned as credits_charged on the completed response — check GET /balance for the account balance.

Common errors for this job type:

`insufficient_credits`	Account balance can't cover the job cost.
`content_policy_violation`	Prompt or media rejected by content policy.
`generation_timeout`	The job exceeded the maximum processing time.
`rate_limited`	Too many requests per minute.
`invalid_input`	A required parameter is missing or outside the model's schema.
`generation_failed`	Generic processing failure.