The method used to update an image during sampling.
Speed depends on the choice Affects detail, speed, and the response to other settings
Samplers differ in speed, detail, and how they respond to step count and noise schedule. Start with the selected model's suggestions.Ancestral samplers, often marked with an _a suffix, add noise during sampling. SDE samplers also use stochastic methods, which involve randomness. That can make the result change noticeably when you change the step count.Other samplers move towards one image as the step count rises, so extra steps refine the picture more than they reshuffle it, though the composition can still shift. The two example sets show one sampler of each kind.Some samplers evaluate the model more than once per step. Check the work-per-step note and kudos estimate when comparing cost.
The examples on this page were made from one prompt, one model and one seed, changing only the setting shown. Other pictures, models and settings respond differently; compare on your own request with the seed and the other settings fixed.