Edit an image with one or more references. Compatible with OpenAI's Images edits API, so the official SDKs work unchanged.

The body is multipart/form-data: reference images go as files, everything else as ordinary fields.

Billing matches the generation endpoint plus one extra component: reference image input. Reference images are billed as input tokens; the request reserves an estimated amount up front and settles against the actual usage in the response.

The same fields can also be sent to /v1/tasks (also as multipart) to use the asynchronous workflow.

Streaming works the same as on the generation endpoint, with the events named image_edit.partial_image / image_edit.completed. Note that ordinary fields in a raw multipart request are strings, so send stream=true as a text field.

{
  "created": 1786000000
}