Halftone

The newsprint and comic look is actually a physical process: continuous-tone images get reproduced as a grid of ink dots whose size grows with tone. Dark spots have big dots, bright spots have small ones. That's halftone screening, and it's surprisingly hard to do convincingly - most "halftone" filters just tile a dot pattern over the image and call it a day. Halftone in the Darkroom pack builds the dots from the actual image tones, with real screen angles and dot shapes, and it's GPU-accelerated to boot.

How it works

The node converts image tones to dot coverage and then renders a dot grid where each dot's size tracks the local tone. But the real details are where it shines:

• Two methods. AM (amplitude modulation, the default) renders clustered dots whose size carries the tone - the classic newsprint look. FM (frequency modulation) uses dispersed Bayer dithering - the stochastic-style "ink jet" pattern where dot spacing varies instead of size. The FM mode ignores dot_shape, angle, and supersample, as the tooltip warns.
• Four dot shapes - round (classic), line (engraving/linocut screen), square (retro), and ellipse (chain dots). The shapes are tone-linearized, so changing the shape changes dot geometry, not overall brightness.
• Real screen angles. Mono defaults to 45° - the classic newspaper angle. In color (CMYK) mode it uses the standard rosette at 15/75/0/45°, the actual angles print shops use.
• lines (20–400) sets screen frequency - lines across the long edge - and it's resolution-independent, so the screen reads the same at any image size.
• black_generation - CMYK only, 0–1. 1 = full GCR (black plate derived from min CMY), 0 = CMY only. It's the same GCR (Gray Component Replacement) control print workflows care about.
• supersample (1–4) anti-aliases dot edges. Higher is smoother and slower.
• strength - blend back to the original. Less than 1 gives a subtle screen overlay, which is a genuinely nice stylize trick.

Mono or color?

color_mode toggles between mono (black) - newsprint black on white, the classic look - and color (CMYK), a naive CMYK separation that gives you the four-color rosette. The tooltip and README are both explicit that this is a stylize effect, not a calibrated proof - if you actually need to know what an image looks like on press, use the pack's CMYK Soft-Proof node instead. But for a comic book cover or a zine aesthetic, it's spot on.

Where it fits

It's under AKURATE/Darkroom/Print, alongside the CMYK workflow nodes. It's a spatial effect (each dot depends on its neighbors), so the README excludes it from LUT bake chains - apply it after extraction. Output is a single image.

One practical note: halftone dots live and die by resolution. If you halftone a small image, the dots are chunky and the print look gets lost; upscale first, then halftone, then downscale if needed. The resolution-independent lines control helps, but dot edges still need pixels to look like dots.

Installing it

Part of the full ComfyUI-Darkroom pack:

cd ComfyUI/custom_nodes
git clone https://github.com/jeremieLouvaert/ComfyUI-Darkroom.git
pip install -r ComfyUI-Darkroom/requirements.txt

Or search "Darkroom" in ComfyUI Manager. Needs scipy, opensimplex, rawpy, exifread, Pillow. Restart and it's under AKURATE/Darkroom/Print. Unlike most of the pack, this one runs its dot rendering on the GPU (torch) - still no model files, no API, nothing to license.