⚡ Quick Answer
KSampler is the ComfyUI node that actually generates your image. It takes your model, positive and negative conditioning, and a starting latent, then runs the denoising process for a set number of steps to produce the finished latent.
Every other node in a basic workflow exists to feed KSampler what it needs — the checkpoint, the prompt, the empty canvas. KSampler is where those pieces actually get combined and the noise-to-image process happens, controlled by widgets like steps, cfg, sampler_name, scheduler, and denoise.
Where You'll See It
Sitting in the middle of nearly every workflow, wired from an Empty Latent Image (or an image-to-image source) on one side, a Load Checkpoint and two CLIP Text Encode nodes for conditioning, and outputting to a VAE Decode node on the other side to turn the result into a viewable image.
Quick Example
A typical SDXL text-to-image KSampler runs at around 25–30 steps, cfg 6–8, sampler_name dpmpp_2m, scheduler karras, and denoise 1.0 for a full generation from an empty latent.
Common Mix-Up
Frequently Asked Questions
See It In Action
Ready to run your first generation?
Our text-to-image tutorial walks through every KSampler setting step by step.
Published: 2026-09-04 · Last updated: 2026-09-04
Join the discussion
Sign in to leave a comment or reply
No comments yet
Be the first to share your thoughts!
