Image Gaussian Blur

Exactly what it says: it blurs an image with a Gaussian kernel. One image in, one blurred image out, with a single knob for how strong. There's no clever twist here - it's the Photoshop "Gaussian Blur" filter as a node, and the value is that you can drop it anywhere in a workflow without round-tripping to an image editor.

Why blur inside a generation graph at all? A few honest reasons. Softening a background plate before compositing so a subject sits in it more naturally. Blurring a mask-derived image to feather a hard edge. Taking the bite off a control image so a preprocessor or a depth map doesn't cling to noise. Building a poor-man's bloom or a depth-of-field look. And the classic trick: blur an image, then use it as a low-frequency base that a later pass refines. Small tool, lots of little jobs.

How it works

A Gaussian blur replaces each pixel with a weighted average of its neighbors, where the weights follow a bell curve - nearby pixels count more, distant ones fade off smoothly. The radius controls how wide that neighborhood is. Bigger radius, wider averaging, softer result. It's the standard, artifact-free way to blur; it doesn't ring or block the way a crude box blur can.

The inputs and outputs that matter

Two inputs, that's it:

• images - the image or batch to blur.
• radius - blur strength, default 1, range 1–100 (integer). At 1 the effect is barely-there; by 10–20 you're into clearly-soft territory; toward 100 it's a heavy smear where only broad shapes survive. Start low and climb - blur radius is very non-linear to the eye, and a little goes further than you'd expect.

Output is a single blurred IMAGE.

How to install it

Comes with the Art Venture pack. ComfyUI Manager → search comfyui-art-venture → Install → restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/sipherxyz/comfyui-art-venture

then pip install -r comfyui-art-venture/requirements.txt and restart. No model download - it's pure image math.

Common issues & troubleshooting

A tiny radius does almost nothing. Right - 1 is a whisper. The scale is non-linear and the low end is subtle. If you want a visible soft focus, you're probably in the 5–20 range depending on image size. Larger images need a larger radius for the same apparent softness, since the blur is measured in pixels and a 2048px image has a lot of them.

I wanted to blur a mask, not an image. This node operates on IMAGE types. If your goal is to feather a MASK for inpainting, you're better off with a mask-blur control - Art Venture's own PrepareImageAndMaskForInpaint has a mask_blur parameter built in, and ComfyUI has native mask-blur nodes. Blurring an image that happens to look like a mask can work, but keep the types straight.

It's slow on huge images at high radius. Wide Gaussian kernels over millions of pixels take real time. If you're only after a soft base, downscale first (Art Venture's ImageScaleDownBy), blur, then scale back up - much cheaper and visually near-identical for a heavy blur.

Node missing after a workflow import. The pack isn't installed - Manager → Install Missing Custom Nodes.