CCTech Lineart Anime Preprocessor ⚡

If your output is anime or manga-style art, you don't want "realistic" lineart - you want the clean, even, inked outlines an anime line drawing is supposed to have. That's exactly the job of the Lineart Anime preprocessor: it takes a photo or render and produces crisp anime-style linework that drops straight into an anime lineart ControlNet. It's the natural companion to the pack's realistic Lineart node: same family, different output style, tuned for a different final look.

Anime lineart is also one of those conditions with real staying power. The SDXL xinsir union lists "anime lineart" as one of its twelve modes, and anime lineart ControlNets are a staple of Illustrious-class and SD 1.5 pipelines. For structure-driven anime work - keeping a character's pose and outline while the generator redraws them - this is the preprocessor you reach for.

How it works

The detector is a pix2pix-style U-Net generator - a paired image-to-image network trained to translate photos into anime line drawings. It's a from-scratch architecture port from comfyui_controlnet_aux (Apache-2.0), real model code rather than a wrapper around another pack or a transformers call. The weights auto-download from HuggingFace on first use into ComfyUI/models/lineart_anime/.

Inputs and outputs that matter

• image - a photo, render, or even a rough sketch you want converted to clean anime linework.
• resolution (default 512, 64–2048) - the internal processing size.

That's the whole input surface - no coarse/fine toggle here, no thresholds. One image in, one line drawing out. This is about as simple as a model-based preprocessor gets: set your resolution, connect the output to an anime lineart ControlNet's control_image, and you're done.

Installing it

It's one of eleven preprocessors in ChrisColeTech/ComfyUI-ControlNet-Nodes, sitting under 🤖 CCTech/Preprocessors alongside its realistic sibling. Install through 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 requirements.txt (huggingface_hub, opencv-python).

Common issues

The only moving part is the first-run checkpoint download into models/lineart_anime/ - give the queue a moment the first time. After that it's remarkably boring, which is the point. The main way people trip is wiring the output into the wrong ControlNet: an anime-lineart preprocessor belongs with an anime lineart ControlNet checkpoint, not a realistic one. If your generated art comes out with wobbling, un-anime-like outlines, check which checkpoint you loaded before you blame the preprocessor. For screenshots with compression noise, running a denoise pass on the source first gives noticeably cleaner linework.