AI image seed: what it is and how to repeat an image

An AI image seed is the number that fixes the random starting point of a generation. In a generator that accepts one, the same seed with the same model, prompt, and settings is meant to give the same picture again, and a new seed gives a new variation. Sume's Image API docs list seed as a "seed for deterministic generation", but no model serves it yet, so a request that sends one returns 400 unsupported_parameter.

Sume facts come from the Image API docs, the Video generation docs, and the image request schema in current code, read on 2026-09-27.

How does an AI image seed work?

A generator that takes a seed uses it to pick its random starting values. Keep the seed and every other input fixed, and those values repeat; change only the seed, and you get a different take on the same prompt. People fix a seed to reproduce a result, or to compare prompt edits against the same starting point.

A seed is not a promise of an identical image, and it only means something for the model and settings it was used with. Change the model or the size, and the same number no longer points to the same picture.

Can I set a seed on Sume's image API?

Not today. seed is an integer field in the POST /v1/images schema, but no image model advertises it, so the request is refused with 400 unsupported_parameter. Video is the same: no v1 video model accepts seed, and each reports seed: false; what is an AI video seed covers that side.

How do I get the same image again without a seed?

Work from the image itself rather than from a number:

• Keep the file. The image you liked is the only exact copy, and results come back as Sume-hosted, signed URLs, so save it in your own storage.
• For a close variant, host that image at a public HTTPS URL, send it as a reference with the same prompt, and name the one change you want. The model redraws the whole image, so check that the rest held.
• To explore, ask for several images at once with n and keep the one you like. Each call is a fresh set.
• For a series with the same person or product, see consistent character AI image generator.

Can I use a negative prompt instead?

No. The image request has no negative_prompt field, and in current code a field the schema does not define is refused with 400 invalid_request. Describe what you want to see instead, such as "a plain white wall" rather than "no posters". Video has its own answer in AI video negative prompt.