A language model sits in front of the generator and fills in what you left out — lens, light, texture, mood. Six words become sixty. For a casual prompt this is usually an upgrade, because the generator behaves better with specifics and most people write far too little.
The cost is that you are no longer the author of the input. Two runs of the same short prompt can be expanded differently, so results vary in ways the seed cannot explain, and a detail you deliberately omitted may be invented and inserted. If you have written a careful, precise prompt, expansion can overwrite exactly the restraint you were going for.
It appears under several names across providers and is frequently on by default. Whether it is on is worth knowing before you spend an afternoon concluding that a model ignores your wording.
In practice
- On by default for many models — check before assuming your text reached the generator verbatim.
- Turn it off when reproducing a result exactly, because the rewrite is a second source of randomness.
- Leave it on for quick exploration from short prompts; it raises the floor considerably.
The mistake to avoid
Debugging a locked seed while expansion is enabled. The seed is doing its job; the prompt going in was different each time.
Related terms
Prompt
A prompt is the written instruction a generative model reads to decide what to make — the one input almost every model requires.
Prompt adherence
Prompt adherence is how faithfully a model does what the prompt actually said, as opposed to producing something attractive in the same neighbourhood.
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.
Negative prompt
A negative prompt lists what you do not want in the output, steering the generation away from those terms instead of towards them.
Aspect ratio
Aspect ratio is the proportion between an output's width and its height, written as two numbers — 16:9 for landscape, 9:16 for a full phone screen.
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync in a single app.