Nodes/SP-Nodes/FaceScatter
ComfyUI Node

FaceScatter

Composite a face across a canvas

By bananasss00·Created 2 years ago·Updated 6 days ago· 20
FaceScatter
  • image
  • mask
  • scatter_params
  • IMAGE
  • MASK
width512
height768

This one does roughly what the name says: hand it an image, a mask, and a scatter_params configuration, and it composites that source across a canvas at whatever size you set. It sits in SP-Nodes alongside the pack's other face-processing pieces - face detection masks, a face processor loader - so the practical use case is almost certainly face-swap or face-reference prep rather than general-purpose layout work, though the brief for this node doesn't spell out the exact intent.

How it works

  • image and mask - the source you're compositing and what part of it counts as the subject.
  • scatter_params - a custom SCATTER_PARAMS type, meaning it's configured by a separate companion node elsewhere in the pack rather than directly on this node's own widgets. This brief doesn't expose what knobs live inside that config, so treat scatter_params as connective tissue you wire in from that other node rather than something you set here.
  • width, height (default 512 × 768) - the size of the output canvas everything gets composited onto.
  • Outputs: IMAGE (the composed canvas) and MASK (presumably marking where the pasted content landed - useful if you're chaining into an inpaint pass right after).

width and height set the canvas independently of whatever size your source image actually is - the node is responsible for fitting the scattered content into that target size, not just passing the input dimensions through. That's consistent with a reference-sheet or dataset-prep use case: you'd typically want every output at the same fixed canvas size regardless of what resolution the source face crops came in at.

Where to look if you want more control

The pack also ships FaceScatter2, which exposes several more knobs directly on the node - transparency, mask inversion, image flipping, and a seed for randomization - none of which are present here. If you're starting fresh, it's worth trying FaceScatter2 first; this original version reads as the simpler, non-randomized predecessor.

How to install it

Through ComfyUI Manager: search "SP-Nodes," install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/bananasss00/ComfyUI-SP-Nodes

Restart ComfyUI afterward.

Common issues

Documentation on this specific node is thin - there's no author description in the schema, and the README doesn't call it out separately from the pack's general "another nodes" catch-all. If scatter_params isn't behaving the way you expect, your fastest path to understanding it is opening the companion config node's own tooltip inside ComfyUI, or checking the pack's example workflows for a working setup you can reverse-engineer. That's an honest gap rather than a specific bug to watch for - this brief simply doesn't have enough information to tell you more about that input than its type.

CategorySP-Nodes

Inputs (5)

NameTypeDefaultDescription
imageIMAGE
maskMASK
scatter_paramsSCATTER_PARAMS
widthINT5121–16384
heightINT7681–16384

Outputs (2)

NameTypeDescription
IMAGEIMAGE
MASKMASK