Edge Enhance

IsekaiEdgeEnhance is a three-mode edge filter: two strengths of edge enhancement and a "Find Edges" mode that extracts line art. The first two are for making a slightly soft image read crisper. The third is the interesting one - it turns a photo or render into a line drawing, which is exactly the kind of preprocessing people feed into ControlNet-style workflows when they want to drive a generation from an outline.

How it works

It's a wrapper over PIL's ImageFilter with three methods, each a different flavor of the same idea (find where neighboring pixels differ, then emphasize that difference):

• Edge Enhance - a mild convolution that boosts local contrast at edges. Subtle; good for a light crispness pass.
• Edge Enhance More - the same idea pushed harder. Visible sharpening, can start looking crunchy on detailed images.
• Find Edges - the big one. It detects edges and renders them as lines on a black background, discarding most of the original image's tone. That's your line-art extraction.

If you've used the ControlNet canny preprocessor, you already know the genre: edge maps are cheap, deterministic structural descriptions of an image, and they're a classic input for structural guidance. Find Edges here gives you a rougher, looser version of that - it's PIL's edge operator, not a trained canny model, so expect thicker, less precise lines. For stylization it's perfect; for precise ControlNet conditioning, a real preprocessor is still the better tool.

The inputs

• image - the image to process.
• method - Edge Enhance, Edge Enhance More, or Find Edges.

That's it - no strength slider, no threshold. Output is a single image tensor.

Installing it

Ships in isekai-comfy-node:

ComfyUI Manager: search "isekai" → install isekai-comfy-node → restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/isekai-sh/isekai-comfy-node
cd isekai-comfy-node
pip install -r requirements.txt

Restart and it's under Isekai → Image/Effects. Pillow and requests; no models, no downloads.

Common issues

Two things to keep in mind. First, the pack-wide batch caveat: batched images only get their first frame processed. Second, Find Edges is binary in a practical sense - there's no threshold control, so noisy or low-contrast images produce messy, dense line maps. If your edge output looks like static, it's not broken, that's just what PIL's operator does with a busy image.

A practical note: for the sharpen-like modes, this node is a blunt instrument compared to a proper unsharp mask. Use Edge Enhance for "make it a touch crisper" - if you want surgical sharpening, a dedicated sharpen node with a radius control will treat you better. Know what each tool is for.