GPT Image 2

GPT Image 2 uses the same model key for generation and editing. Omit image_urls for text-to-image, or provide up to 10 public HTTP(S) images for editing.

Request parameters

Parameter	Required	Accepted values
`prompt`	Yes	Non-empty UTF-8 text, maximum 32,000 bytes at the gateway
`image_urls`	No	1–10 public HTTP(S) image URLs
`size`	No	`auto`, `1024x1024`, `1536x1024`, `1024x1536`
`quality`	No	`auto`, `low`, `medium`, `high`
`output_format`	No	`png`, `jpeg`, `webp`

{
  "prompt": "A cinematic product photo of an orange mechanical keyboard",
  "size": "1536x1024",
  "quality": "high",
  "output_format": "png"
}

Submit to POST /api/v1/task/submit/gpt-image-2. Save the returned data.task_id and poll /api/v1/task/status.

Each request produces one image. The current gateway price is fixed regardless of the optional parameters, but /pricing remains the current source of truth.

See the GPT Image 2 API reference for the machine-readable Schema.

Getting started

Agent integrations

Video Generation

Image Generation

GPT Image 2

GPT Image 2

Request parameters

​GPT Image 2

​Request parameters

GPT Image 2

Request parameters