AI Glossary

Sampling steps

Also called inference steps, denoising steps

What does “Sampling steps” mean?

Sampling steps: Number of rounds a diffusion model uses to turn random noise into a finished image. More steps usually add refinement up to a point, then give diminishing returns while taking longer. Distilled "fast" models are designed to produce good results in only a few steps.

Sampling steps visual guide: Number of rounds a diffusion model uses to turn random noise into a finished image. More steps usually add refinement up to a point, then give diminishing returns while taking longer. Distilled "fast" models are designed to produce good results in only a few steps.
Visual guide Open full size (new tab)

What is an example of “Sampling steps”?

Raising steps from 4 to 30 on a standard model may sharpen details, but going from 30 to 100 often changes little except generation time.

Keep learning

See also

SamplerTermCFG scaleTermSeedTermDiffusion modelTerm

Same topic

Denoising strengthTermSpeech-to-textTermOutpaintingTermStyle transferTermNeRFTerm