GPT Image revised_prompt: what a Sume images response returns

OpenAI returns the rewritten prompt in a revised_prompt field on the image generation call. A Sume POST /v1/images response has no such field: it returns data[].url, media_type and a usage object, so log the prompt you sent if you need to know what produced an image.

OpenAI facts are from its image generation guide; Sume facts from Image generation, both read 2026-10-01.

Where does OpenAI put the revised prompt?

In the guide's "Revised prompt" section, the mainline model revises your prompt when you use the image generation tool in the Responses API. You read the result from the revised_prompt field of the image generation call. The guide's example shows a descriptive sentence about a gray tabby cat hugging an otter.

What does a Sume images response contain?

The Sume docs say images "return as Sume-hosted URLs with usage data". The fields are below.

Does Sume rewrite my prompt at all?

The docs describe prompt as a text description of the desired image and list no field that reports a rewritten version. They do not state whether any model rewrites the prompt internally, so do not assume either way. What the response gives you is the generated image URL and the billed cost.

How do I keep an audit trail of prompts?

Store the prompt string next to the returned data[].url in your own records. The request also accepts a metadata object that the docs say is stored on the job and not sent to the provider, which is a place to keep your own prompt id or template name. For related behavior on long requests, see how a long image request becomes a job.