Image Canny Filter

Canny edge detection is the classic - it finds hard, well-defined edges in an image and hands you back a clean, mostly black-and-white line map. If you've done any ControlNet work you already know Canny as the go-to preprocessor for architecture, product shots, and anything with clear, mechanical contours: it produces thin, confident lines rather than the softer gradient edges you'd get from something like Softedge. WAS Node Suite ships its own implementation of that same algorithm, so if you're already running WAS for other things, you get a Canny node for free without pulling in a second, dedicated preprocessor pack.

What it's actually for

Two real use cases. The first, and the one most people land here for, is ControlNet conditioning: run your source image through this node, wire the resulting edge map into a ControlNet Apply node alongside a Canny ControlNet model, and you get generation that respects the original composition's structure. The second, less common but genuinely useful, is just wanting an edge map on its own - inspecting where the hard boundaries in an image actually are, building a stylized line-art look, or using edges as a mask input elsewhere in a workflow.

Worth being upfront about which pack most ControlNet workflows actually reach for: the dedicated ControlNet Auxiliary Preprocessors pack is the community-standard place to get Canny, depth, pose, and the rest of the preprocessing family in one place, and most shared workflows assume it. WAS's Canny filter is a fine substitute if you already have the whole WAS suite installed and don't want a second dependency just for one preprocessing step - but if you're setting up ControlNet from scratch, the Auxiliary Preprocessors pack is still the one most tutorials and shared workflows expect you to have.

How it works

Same idea as any Canny implementation: it looks for sharp gradients in brightness - places where pixel values jump quickly - and traces those into thin, connected lines while suppressing everything softer. That's why it's specifically good at hard-edged, high-contrast content and less good at organic, painterly subjects, where "an edge" is a much fuzzier concept and the filter tends to pick up noise and texture as spurious lines rather than shape. You feed it an image and get an image back - the edge map, same dimensions as your source.

Installing it

Bundled with the whole pack:

• ComfyUI Manager - search WAS Node Suite, install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/WASasquatch/was-node-suite-comfyui/, then pip install -r requirements.txt from inside that folder, then restart ComfyUI.

Common issues

If your edge map comes out either far too noisy (every bit of skin texture and fabric grain showing up as an "edge") or far too sparse (missing shapes you'd expect to see), that's Canny doing exactly what Canny does on the wrong kind of source image - it's tuned for hard-edged content, not soft or organic scenes. On portraits or painterly images, a Softedge or Lineart-style preprocessor from the Auxiliary Preprocessors pack will usually give you a more usable result for ControlNet than pushing Canny to work on something it wasn't built for.

The pack-wide issue is the one to know going in regardless of which node you're using: WAS Node Suite has been unmaintained since December 2023, and "Import Failed" errors after a ComfyUI update are a recurring, real complaint - the pack's pinned dependencies can drift out from under a newer ComfyUI install. Reinstalling requirements.txt inside your ComfyUI venv is the usual fix if the whole pack stops loading.