CCTech Manga Line Preprocessor ⚡

The lineart family in this pack is really three different aesthetics. Realistic lineart gives you natural illustration strokes, anime lineart gives you clean inked outlines, and this node - Manga Line - is the specialist: it extracts the kind of crisp, high-contrast linework you'd associate with finished manga pages, where the ink does the talking and screentone does the shading. If your ControlNet pipeline is aimed at manga-style output, this is the preprocessor that speaks its language.

Manga line extraction comes from a lineage of dedicated line-extraction research - the sort of model built to pull clean line drawings out of noisy photographs so they can be re-inked. That's a different target from a general edge detector, and it shows: where Canny and HED chase every boundary, this one pushes toward drawing-quality linework.

How it works

The detector is built on the res_skip CNN from the MangaLineExtraction project - a residual-skip architecture that's good at preserving clean line structure while suppressing shading and texture noise. As with every node in this pack, it's a from-scratch port of the architecture from comfyui_controlnet_aux (Apache-2.0) rather than a wrapper. The weights download from HuggingFace on first use into ComfyUI/models/manga_line/.

Inputs and outputs that matter

• image - a photo or render you want converted to manga-style linework.
• resolution (default 512, 64–2048) - the internal working size.

No thresholds, no toggles, no model switcher. It's the second-simplest node in the pack after Canny: image in, line image out. Connect the output to the control_image of a lineart-class ControlNet - manga linework feeds the same lineart-conditioned checkpoints as the other line extractors - or use it as a structural input for an edit model.

Installing it

It ships in ChrisColeTech/ComfyUI-ControlNet-Nodes, the eleven-preprocessor pack under 🤖 CCTech/Preprocessors. Install via ComfyUI Manager (search "ComfyUI-ControlNet-Nodes") or:

cd ComfyUI/custom_nodes
git clone https://github.com/ChrisColeTech/ComfyUI-ControlNet-Nodes

Restart ComfyUI. No dependencies beyond the pack's own huggingface_hub and opencv-python.

Common issues

First run downloads the checkpoint - expect a pause, then smooth sailing. The thing to keep in mind is scope: this is a specialist, not a general-purpose edge detector. On subjects without strong ink-like contrast it can look sparse compared to Canny or HED, which isn't a bug - it's refusing to ink every wrinkle. If you want structure everywhere, use a broader preprocessor; if you want manga-grade linework, this is the one. Matching it to a lineart ControlNet checkpoint (rather than, say, a canny one) is the difference between workable and fighting your own condition.