Image Lucy Sharpen

Most "sharpen" nodes are a glorified unsharp mask - they find edges and crank the local contrast, which looks sharp until you notice the ugly halos around everything. Image Lucy Sharpen takes a different, smarter route: Richardson–Lucy deconvolution. Instead of just exaggerating edges, it tries to reverse the blur - it models the image as a sharp original that got smeared, and iteratively works backward toward that original. Done right, you get recovered detail with far fewer of the crunchy halos.

Why you'd use it

For gently rescuing softness. AI generations, upscales, and downsized images often come out a touch mushy, and this can bring back a sense of crispness that looks more like real detail than like a sharpen filter. It's the one to reach for when a plain sharpen makes things look processed and edgy and you want something that reads as cleaner rather than sharper. It's a post-processing step - run it near the end of a workflow, after generation and any upscale.

How it works, and the knob that matters

The core control is iterations - how many deconvolution passes it runs. This is the whole game:

• Too few and you barely change the image.
• A sensible handful and you get that nice recovered-detail look.
• Too many and it overshoots badly: noise gets amplified, edges start ringing, and grain in flat areas blooms into something worse than the blur you started with.

Deconvolution amplifies whatever's there, signal and noise, so more is emphatically not better. Creep the iterations up from a low value and stop the moment it looks right - the failure mode is always "I pushed it too far." It's also more expensive than a simple sharpen because it's iterating, so on big images it's not instant.

The input is an image and the output is the sharpened image, ready to save or carry on processing. Because it can amplify noise, it plays best on reasonably clean input - if the source is grainy, denoise a little first, then sharpen.

Installing it

Ships in WAS Node Suite. Install the pack via ComfyUI Manager (search was-node-suite-comfyui, install, restart) or clone it:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui

install requirements.txt, restart. No model files - it's an image-math operation.

Troubleshooting

If the result looks noisy or ringy, that's not a bug - it's too many iterations. Back them off. The pack-wide problem to know about: WAS Node Suite has been unmaintained since December 2023, and the whole suite can throw "Import Failed" after a ComfyUI update from a pinned-dependency clash (opencv is the usual culprit, and the imaging nodes lean on it). Reinstall requirements.txt into the correct venv (activate it, or use install.bat) to reconcile it.