Nodes/ComfyUI Fooocus Nodes/Fooocus KSampler
ComfyUI Node

Fooocus KSampler

The simplest node in the pack, and the easiest to forget

By Seedsa·Created 3 years ago·Updated 2 years ago· 102
Fooocus KSampler
  • pipe
  • model
  • pipe
  • image
image_outputPreview
save_prefixComfyUI

By the time you're here, all the interesting work is done. The Fooocus PreKSampler prepared the conditioning, applied the styles, handled the inpaint patch and the ADM scalers. This node is the actual denoising pass - the thing that turns all that preparation into pixels - and then it hands you the image. It's the output node of the pack's whole text-to-image flow, and it has one input that matters.

What it does

It takes the prepared pipe and samples it. That's it. Where a standard KSampler wants you to hand it model, positive, negative, latent and a sampler/scheduler pair, this one takes a single pipe and pulls all of that out of it. The settings you'd normally fight with on a KSampler were already decided upstream - mostly at the PreKSampler - so this node stays deliberately boring.

The inputs:

  • pipe - the prepared pipeline. Required.
  • image_output - Hide, Preview, Save, or Hide/Save. Preview is default and sensible: it shows you the result without committing it to disk.
  • save_prefix - the filename prefix when you do save.
  • model (optional) - override the model if you've patched one in elsewhere, e.g. after an ApplyImagePrompt.

Outputs: pipe (with the sampled result and latent), and image - the actual picture. Wire the image to a Save node or straight on into the Fooocus Upscale / Fooocus detailerFix chain.

Installing

Same pack install as always - ComfyUI Manager, search "Fooocus Nodes", or clone the repo into custom_nodes and restart. No model download here; everything heavy happened upstream.

Where people get confused

The classic failure: you wire a Loader to a PreKSampler, stare at the graph, and wonder why nothing renders. The PreKSampler prepares - it doesn't sample. This node is the thing that actually produces the image, and if it's missing, your workflow runs and quietly outputs nothing useful. The second confusion is the opposite direction: people try to use this as a drop-in replacement for a standard KSampler in a normal ComfyUI graph. It won't accept the individual model/conditioning wires - it wants a pipe. If you need to break out, that's what the Fooocus PipeOut node is for.

CategoryFooocus

Inputs (4)

NameTypeDefaultDescription
pipePIPE_LINE
image_outputCOMBOPreview4 options: Hide, Preview, Save, Hide/Save
save_prefixSTRINGComfyUI
modeloptMODEL

Outputs (2)

NameTypeDescription
pipePIPE_LINE
imageIMAGE