What is a sampler in AI image generation?

The algorithm that steps the model from random noise to a finished image, deciding the size and shape of each denoising step.

Which sampler is best?

For most work, any modern deterministic one. Differences between current samplers affect fine detail rather than composition or style, and the choice matters far less than the prompt.

What does the "a" in Euler a mean?

Ancestral. It injects fresh noise at every step, so the image keeps changing as steps increase instead of converging on a stable result.

How many steps should I use?

Enough to converge and no more — often 20 to 30 with a higher-order sampler. Past that point extra steps cost time and change nothing you can see.

Why do I get a different image when I change sampler?

The same starting noise is being walked to the result along a different path, so the fine detail lands differently even though the prompt and seed are unchanged.

Does the sampler affect image quality?

Marginally, and mostly through how many steps are needed. It does not change what is in the picture.

Why does my tool not have a sampler setting?

Many hosted models do not expose one, having chosen a sensible default. Given how little it changes, that is a defensible decision rather than a missing feature.

AI image glossary

What is a sampler?

The algorithm that steps the denoising process. Different samplers converge at different speeds and produce slightly different results from the same seed.

A sampler is the algorithm that walks a diffusion model from random noise to a finished image, deciding how large each denoising step is and how it is taken. Different samplers reach a usable result in different numbers of steps and produce slightly different images from the same seed and prompt, though the differences between modern samplers are far smaller than the interface implies.

Stepped paper swatches climbing from dark to light beside a fine brush on warm paper

What the sampler is actually doing

The model can predict how much noise is present in an image at a given moment. It cannot, in one move, turn static into a photograph. Getting there requires many small steps, and something has to decide how big each step is, whether to look ahead, and how to correct for the error introduced by moving in discrete jumps rather than continuously. That is the sampler. It is a numerical solver, and the names in the dropdown are the names of solvers.

Reading the names in the list

Euler is the simplest and does what its name suggests: step forward by the predicted amount and repeat. The "a" in Euler a marks an ancestral sampler, which injects fresh noise at each step and therefore never fully settles — it keeps producing variation as steps increase rather than converging. DPM++ and its relatives are higher-order solvers that reach a comparable result in fewer steps by estimating the curvature of the path rather than assuming it is straight. UniPC and similar newer entries are further refinements of the same idea.

Ancestral versus deterministic, which is the difference that matters

Of everything in that dropdown, this is the distinction with practical consequences. A deterministic sampler converges: raise the step count and the image stabilises, so 30 steps and 50 steps give nearly the same picture. An ancestral sampler adds noise every step, so the image keeps changing as you add steps and never lands anywhere final. If you are trying to reproduce an exact result or tune one image, ancestral samplers will frustrate you.

Steps, and the point where more stops helping

Sampler choice and step count are one decision rather than two. A higher-order sampler may produce at 20 steps what a simple one needs 40 to reach, and beyond the point of convergence extra steps cost time and change nothing visible. Doubling the steps when an image is not working almost never fixes it; the prompt, the seed or the model is what is wrong.

Why most people should leave it alone

Sampler choice is the most over-discussed control in image generation and one of the least consequential. It changes fine texture and the exact arrangement of small details, not composition, subject or style. Anyone whose images are not working will get far more from rewriting the prompt than from moving down this list, and many current hosted models do not expose the setting at all — which is a reasonable reading of how much it matters.

Sampler at a glance

Is: A numerical solver
Decides: How each step is taken
Ancestral: Never fully converges
Deterministic: Settles as steps rise
Typical steps: 20 to 30
Effect on composition: None

Related terms

What a sampler sits next to

reproducing an imageThe noise it starts from.

What is guidance scale?How hard the prompt pulls.

What is a checkpoint?The weights it steps through.

What is Stable Diffusion?Where these names come from.

build a promptThe control that actually matters.

terms explainedEvery term, defined.

Put it to use

Where this appears

generate from a descriptionThe main generator — text to finished image.Open 120 images and their prompts120 images with the exact prompts that made them.Open start from a photoStart from a picture you already have.Open AI Photo EditorChange part of a photo, keep the rest.Open enlarge an imageEnlarge to 4K without the smearing.Open AI Photo EnhancerExposure, noise and old-photo restoration.Open remove something from a photoMasked removal that rebuilds the background.Open depth-aware blurDepth-aware, not a uniform filter.Open add colour to an old photoPlausible colour for black and white.Open AI Logo GeneratorMarks that still read at 16 pixels.Open design a tattooDesigns that survive being tattooed.Open generate illustrationIllustration and painting, not photographs.Open generate a headshotStudio-grade portraits without a studio.Open Passport PhotoThe exact specification, and how to meet it.Open AI Interior DesignRestyle a room you can photograph.Open AI Image GlossarySeed, denoise, inpainting, explained.Open

Sampler — questions people ask

What is a sampler in AI image generation?: The algorithm that steps the model from random noise to a finished image, deciding the size and shape of each denoising step.
Which sampler is best?: For most work, any modern deterministic one. Differences between current samplers affect fine detail rather than composition or style, and the choice matters far less than the prompt.
What does the "a" in Euler a mean?: Ancestral. It injects fresh noise at every step, so the image keeps changing as steps increase instead of converging on a stable result.
How many steps should I use?: Enough to converge and no more — often 20 to 30 with a higher-order sampler. Past that point extra steps cost time and change nothing you can see.
Why do I get a different image when I change sampler?: The same starting noise is being walked to the result along a different path, so the fine detail lands differently even though the prompt and seed are unchanged.
Does the sampler affect image quality?: Marginally, and mostly through how many steps are needed. It does not change what is in the picture.
Why does my tool not have a sampler setting?: Many hosted models do not expose one, having chosen a sensible default. Given how little it changes, that is a defensible decision rather than a missing feature.

Related AI image terms

Start creating today

Generate your first image with the best AI image generator.

Turn a sentence into a finished, photorealistic image in seconds — then refine every detail. No setup, no Discord, no GPU.

Generate without picking a sampler Get Started

Join 4,200+ creators using LaFoto