Generation controls

    Seed

    Also called Random seed, Noise seed.

    A seed is the number that decides the random starting noise for a generation, so the same seed with the same settings reproduces the same output.

    Generation starts from random static. The seed is what makes that static repeatable: it is the number fed to the random-number generator, so seed 42 produces one particular field of noise and seed 43 produces a completely different one. Everything downstream is deterministic, which is why identical inputs and an identical seed give you the identical result.

    That makes it the only honest way to A/B a prompt. Change a word with the seed floating and you have changed two things — the wording and the starting noise — so you cannot attribute the difference. Lock the seed and the wording is the only variable left.

    Leaving it blank is usually the right default. Most tools then pick one at random per run, which is what you want while you are exploring; the moment a result is nearly right, capture that seed and start iterating from it.

    In practice

    • Same seed, same prompt, same model, same settings — same output. Change any one and the guarantee is void.
    • Seeds do not transfer between models. Seed 42 on one model has nothing in common with seed 42 on another.
    • Neighbouring seeds are not related: 41 and 42 are as different as 41 and 900,000.

    The mistake to avoid

    Assuming a good seed carries a good result across model versions. A provider update changes the mapping from noise to picture, and yesterday's seed is just a number again.

    Where you will run into it

    Related terms

    The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync in a single app.