Download Image Content

This is the endpoint the url values in a generation or edit response point at. You do not normally construct the URL yourself — take it from the response.

The URL is authenticated, not public. Fetching it carries the same credential as every other endpoint on this API, and only the organization that generated the image can read it. From your own code that means your API key. A URL pasted into a browser address bar, or dropped into an <img src>, will not load: neither carries an origin we recognise, so the request is refused before it reaches the image.

Images live for 24 hours. After that this endpoint returns 404 and the image is unreachable. There is no delete endpoint — expiry is the only removal path, and it is why url delivery is not a storage product. Requests made with the default response_format: "b64_json" store nothing at all and have no URL to fetch.

image_id is the UUID carried in the url field of the generation or edit response. Every way of missing — an unknown id, a malformed one, an image that has expired, and one belonging to another organization — returns the same 404, so a 404 never confirms that an id exists.

HEAD is supported and answers the image's exact byte length; the GET body is streamed and therefore chunked, so it carries no Content-Length.

Because the body streams, the 200 is committed before the first byte is read. A storage failure therefore shortens the body rather than changing the status: you can receive a truncated image, or in principle an empty one, under a 200. HEAD answers from the recorded length rather than from storage, so comparing the bytes you received against it catches both — worth doing if a partial image would be a problem for you.

Range requests are not supported and not advertised — a Range header is ignored and the whole image comes back with 200, rather than a 416.

Downloading is free and unmetered. This endpoint is authenticated but not rate-limited, not credit-gated and not billed: you were charged for the image when it was generated, so fetching it — or retrying a failed fetch — costs nothing and consumes none of your request quota. An organization with a zero credit balance can still download images it has already paid for.