ImageFilterEdgeEnhanceMore

If your image reads a little soft - text edges, line art, small detail - and you want to firm it up without running a sharpen filter that leaves halos everywhere, this is the node. ImageFilterEdgeEnhanceMore is the stronger sibling of the plain edge-enhance filter: it bumps the contrast at high-frequency transitions so contours pop, and it's aggressive enough that on an already-sharp photo you'll notice it. Think of it as a "definition" button rather than a fix for blur.

It's one of the filter nodes from the Allor Plugin (Nourepide's ~90-node image-processing pack), and it's dead simple: one IMAGE in, one IMAGE out, no parameters to argue with. Wire it between your VAE decode and whatever comes next - upscaler, SaveImage, or a composite - and tweak by adding it after a mild blur for a fake "clarity" pass. It's also handy as a cheap pre-step for feeding edges into a ControlNet preprocessor, though you'll usually want the dedicated FindEdges node for that job.

How it works

Under the hood this is Pillow's ImageFilter.EDGE_ENHANCE_MORE, applied per image. Pillow's edge-enhance family works by detecting edges (high local gradients) and scaling up their contribution, so bright detail gets brighter and transitions snap. MORE is exactly the same operation as plain edge enhance, just with a stronger multiplier. That's why there's no size or strength input - you get one fixed behavior, and you tune the overall effect by how many times you pass the image through, or by what you blur first.

One thing worth knowing: Allor routes it through the PIL path (tensor → PIL image → filter → tensor), and it handles batches and RGBA transparent images fine. It's not the fastest filter in the pack - the cv2-based blurs are quicker - but for a single image it's instant.

Wiring it up

Just the essentials:

• images (IMAGE, required) - the only input. Any ComfyUI image tensor, batched or single.
• IMAGE (output) - the filtered result, same shape and channel count as what you put in.

That's the whole schema. Drop it inline and let it pass through.

Installing it

It ships inside the Allor pack, so you install the pack once and this node appears. In ComfyUI Manager, search "Allor Plugin", or:

cd ComfyUI/custom_nodes
git clone https://github.com/Nourepide/ComfyUI-Allor

then restart ComfyUI. The pack's install script (install.bat / install.sh) activates your venv and runs requirements.txt, which pulls rembg, onnx and gitpython - yes, you get the whole background-removal stack even if you only wanted filters. On first start it also writes a config.json into the pack folder; you can disable modules you don't use there.

Gotchas

Two things bite people on this pack. First, the repo was rebased to strip image history, and the auto-updater (on by default, daily) can throw when it does a soft git pull. If you see update errors on startup, set update_mode to hard or update_frequency to never in config.json. Second, because the effect is fixed-strength, don't stack this node several times hoping for a subtle look - you'll overshoot straight into crunchy territory. Once is usually plenty; pair a mild Gaussian blur before it if you want a softer-clarity vibe instead.