Nodes/WAS Node Suite v3/Image High Pass Filter
ComfyUI Node Runs on cloud

Image High Pass Filter

Pull the fine detail out of an image, Photoshop-sharpen style

By WASasquatch·Created 3 years ago·Updated 5 days ago· 1,844
Image High Pass Filter
  • images
  • images
radius10
strength1.5
color_outputtrue
neutral_backgroundtrue

If you've ever used Photoshop's High Pass filter as a sharpening trick, this is the same idea in node form. Image High Pass Filter strips out the low-frequency information from an image - the broad tones, gradients, and smooth areas - and leaves you with just the high-frequency detail: edges, fine texture, grain. On its own the output looks like flat mid-gray with faint edge outlines; the point isn't to look at it directly, it's to use it as an ingredient.

How it works

The mechanism is the same one photo editors have used for decades: blur a copy of the image, then subtract that blurred version from the original. What's left over is exactly the fine detail the blur smoothed away - literally the high-frequency band of the image, hence the name. The larger the blur radius used internally, the coarser the detail band you pull out; a small radius grabs only the finest texture and grain, a larger one starts pulling in edges and shapes too.

Why you'd reach for it

Two common uses. First, sharpening: extract the high-frequency detail, then blend it back over the original using an Overlay or Soft Light blend mode (WAS Node Suite's own Image Blending Mode node handles that) to punch up perceived sharpness without the halo-y look a naive unsharp-mask can produce. Second, texture and detail analysis or transfer - since the output is just the fine detail with none of the color or lighting information, it's a useful building block if you're isolating skin texture, fabric grain, or noise patterns to reuse elsewhere in a graph.

Installing it

Through ComfyUI Manager: search "WAS Node Suite", install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui
cd was-node-suite-comfyui
pip install -r requirements.txt

Pure image math - no model downloads, no heavy dependencies specific to this node.

Troubleshooting

If the output looks almost entirely flat gray with no visible detail at all, that usually means the source image itself is soft - this node can only pull out detail that's actually there. Feeding it a heavily-denoised or already-blurred image (post-upscale with a strong generative upscaler, for instance) will give you a disappointingly empty result; it's not the node failing, it's genuinely nothing left to extract.

Beyond that, the standard WAS Node Suite caution applies: the pack has had no active development since December 2023, so if this node - or the whole suite - disappears from your menu after updating ComfyUI, check your console for an import failure before assuming the filter logic itself is broken. It's almost always a dependency version mismatch elsewhere in the pack, not this node.

CategoryWAS Suite/Image/Filter

Inputs (5)

NameTypeDefaultDescription
imagesIMAGEThe image to extract detail from. A batch is handled image by image.
radiusINT101–500Detail finer than this many pixels is kept and everything broader is discarded. 2 keeps only grain and pores, 10 keeps skin and fabric texture, 100 keeps most of the picture.
strengthFLOAT1.50–255How far the extracted detail is amplified. 1.0 is the raw difference, which is very dark; 1.5 is the usual working level; 10 is extreme and clips most of it to white. 0.0 gives a flat result.
color_outputBOOLEANtrueKeep the detail in colour, or average it to grey. Grey is the safer choice when the layer is going to be blended back for sharpening, since coloured detail can tint the result.
neutral_backgroundBOOLEANtruePut the detail on mid grey instead of black. Grey is what an overlay or soft-light blend expects, because mid grey leaves the layer beneath unchanged; black gives an add-style layer.

Outputs (1)

NameTypeDescription
imagesIMAGEThe extracted detail, as an RGB image the size of the source.