Scheduler (sampler)
A scheduler (or sampler) is the diffusion-model algorithm that controls how noise is removed step-by-step from random noise into a final image — DPM++ 2M Karras, Euler, DDIM, FLOW-Match are 2026 popular choices.
Diffusion models start from random noise and remove noise step-by-step over N inference steps; the scheduler decides how much noise to remove per step + how to sample. Different schedulers trade speed vs quality: Euler (simple, fast), DDIM (deterministic, good for img2img), DPM++ 2M Karras (high quality, default in many UIs), UniPC (fast + quality), FLOW-Match (Flux models). Step counts: 20-30 typical for quality, < 10 for speed (Flux Schnell uses 4 steps). For non-SD models (Flux, Stable Diffusion 3, Pixart) the scheduler may be baked in; for ComfyUI / A1111 users it's a knob. Picking the right scheduler is the difference between 'fast' and 'good' inference.
When to use scheduler (sampler)
- Tuning quality / speed trade-off in SD-family models.
Common mistakes
- Sticking with the default — workload-tuned schedulers can be 2× faster at same quality.
- Using 50+ steps when 20 is identical — wasted compute.
FAQ
What is scheduler (sampler)?
A scheduler (or sampler) is the diffusion-model algorithm that controls how noise is removed step-by-step from random noise into a final image — DPM++ 2M Karras, Euler, DDIM, FLOW-Match are 2026 popular choices.
When should I use scheduler (sampler)?
Tuning quality / speed trade-off in SD-family models.
What are the most common mistakes with scheduler (sampler)?
Sticking with the default — workload-tuned schedulers can be 2× faster at same quality. Using 50+ steps when 20 is identical — wasted compute.
Related terms
- Diffusion model — A diffusion model is a generative neural network that creates images, video, or audio by iteratively denoising random noise toward a learned target distribution.
- CFG scale (classifier-free guidance) — CFG scale controls how strongly a diffusion image model follows its text prompt — higher values stick closer to the prompt, lower values explore more.
- Seed — A seed is an integer that initializes the random number generator inside an image, video, or audio model, making generation reproducible.
Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/scheduler.md.