Image Intensity

Image Intensity is the twin of Image Luminance: it reduces an image to a brightness map that a Recolor ControlNet uses to repaint the scene. Feed it a photo, keep the light and shadow, drop all the color, and let your prompt decide the new palette - colorize a grayscale shot, or completely re-hue a color one without disturbing the shading. Same job as Luminance, slightly different math. For recoloring work these two are the pack's dedicated preprocessors, and which one wins is often just a matter of eyeballing both.

How it works

The node produces an intensity map - a brightness read of the image - that preserves its tonal structure. That map is a ControlNet condition. Wire it into a recolor ControlNet and the model generates a full-color result that follows the brightness pattern exactly while sourcing every color from your prompt. The README maps this preprocessor to recolor_intensity; on the actual recolor ControlNet model it's honest enough to shrug ("Idk. Maybe same as above?") about whether it wants a different checkpoint than the luminance variant, so in practice you point both at the same recolor ControlNet and pick whichever grayscale looks better.

The difference from Luminance is the weighting. Luminance uses perceptual weighting - matching how bright a color appears to the eye - while Intensity is a plainer brightness measure. On many images the results are near-identical; on saturated or unusual color images they diverge, and that's exactly when it's worth trying both.

The inputs that matter

It takes an image and returns an IMAGE - the intensity map - for your ControlNet Apply node. One knob worth touching:

• gamma_correction (default 1, range 0.1–2) - reshapes the brightness curve. Under 1 brightens and flattens the midtones; over 1 deepens the shadows and adds contrast. Adjust it if the recolor comes out too flat or too dark; leave it at 1 for a straight pass.

resolution (default 512) sets the working size.

Installing it

ComfyUI Manager: search ComfyUI's ControlNet Auxiliary Preprocessors, install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/Fannovel16/comfyui_controlnet_aux
pip install -r requirements.txt

Restart ComfyUI. No model download - computing intensity is pure image math, so it's ready as soon as the pack loads. The recolor ControlNet itself is a separate download you'll need for the conditioning to do anything.

Where people get burned

Same traps as its twin. It only makes the grayscale hint - the actual recoloring happens in the recolor ControlNet, which is an SD 1.5 / SDXL-era model, so run these workflows on those bases. If your output is muddy or washed out, reach for gamma_correction and the ControlNet strength dial before suspecting the preprocessor. And don't overthink Intensity vs Luminance: they're two flavors of the same idea, so try both and keep whichever grayscale gives the cleaner recolor. As always, no recolor ControlNet loaded means this map changes nothing.