lineArt2LineArt Assistant

The other three nodes in the ComfyUI-Img2DrawingAssistants pack create sketches and line art from photos. This one, the lineArt2LineArt Assistant, assumes you already have line art - from the Img2LineArt Assistant, from a ControlNet lineart preprocessor, from a scan - and gives it a cleanup and a recolor. It's the pack's finishing pass, and the README's own suggested workflow is exactly that: hang this node off a lineart preprocessor and repaint the result however you like.

If you've ever pulled line art out of a preprocessor and been stuck with white lines on a black background, or wanted navy-blue ink on white paper without doing it in an image editor, this is the node. Model-free again - pure OpenCV on CPU, instant.

How it works

Nothing is extracted here; the input line art is treated as-is. The node takes the image, optionally inverts it (invert, default on - many preprocessors output white lines on black, and this flips that to black-on-white for processing), cleans up stray specks with clean_up (0–5, morphological noise removal), then re-inks the dark strokes in your chosen line_color and repaints the background with bg_color / bg_light. The details slider (25–255, default 240) is the threshold that decides which pixels count as "line" vs. "paper."

The slightly confusing part is the second invert_default toggle (also default on). It controls the other output: the "default" output is an untouched copy of the input, and invert_default decides whether that passthrough copy gets inverted too. So you get your recolored lineArt output plus a normalized copy of the original. If both toggles are on and you feed in white-on-black line art, the "default" output is your original flipped to black-on-white - useful if you want the clean source and the recolored version.

The inputs that matter

• invert (default true) - normalize white-on-black input before processing. Leave it on unless your input is already black-on-white.
• line_color + color_strength (1–10, default 8) - the ink color and its saturation; strength 10 is full color, lower values wash toward white.
• bg_color + bg_light - the paper color; bg_light above 10 washes toward white.
• clean_up (0–5) - speckle removal on the incoming line art.

Outputs

lineArt - the cleaned, recolored line art; wire this to Preview or Save Image. default - the (optionally inverted) untouched copy of your input. The naming is unhelpful, but once you know "default = the passthrough," it's easy to ignore.

Install

ComfyUI Manager, search ComfyUI-Img2DrawingAssistants, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/Isi-dev/ComfyUI-Img2DrawingAssistants

No model files to download. Dependencies: torch, numpy, opencv-python - torch and numpy come with ComfyUI, and Manager installs cv2 for you. One pack-wide gotcha: this pack requests the full opencv-python build while ComfyUI usually ships opencv-python-headless; both provide cv2, so it mostly works, but pip can churn on which package owns the module.

Troubleshooting

Because the input is pre-existing line art, the single biggest failure mode is getting the inversion wrong: feed it black-on-white art with invert on and your lines vanish into the background. If the output looks empty, flip invert. If lines look thin and washed, raise color_strength toward 10. If your line art came from a messy threshold, clean_up 2–3 handles most speckle.

The author's note applies here as much as anywhere: quality in, quality out. Garbage line art in, cleaner-but-still-garbage line art out. But for taking a usable extraction and turning it into colored, polished ink in one click, it's the one node in the pack that does something the others can't.