👾 Histogram View

A washed-out render doesn't look washed out until you compare it to a good one - but its histogram betrays it instantly. Histogram View plots the brightness distribution of your image inside the node, in RGB and/or luminance, so you can see at a glance whether the blacks are clipped, the highlights are blown, or the whole thing is camped in the middle. It's the diagnostic your eyes are too lazy to be.

What you're reading

The histogram counts how many pixels sit at each brightness level, from pure black (left) to pure white (right). Three patterns cover most diagnoses:

• Piled against the left edge - crushed shadows, detail being lost to black.
• Piled against the right edge - blown highlights.
• A hump hugging the middle with empty ends - the classic "generated image looks flat" profile. This is what Auto Levels from the same pack exists to fix; the histogram is how you confirm it's needed, and how you confirm it worked.

The controls

• show_rgb (default on) - overlays the red, green, and blue channels separately. The channel split is how you spot a color cast: if the red trace sits left of the blue trace across the tonal range, your image is blue-dominant, and vice versa.
• show_luminance (default on) - the single brightness trace using luma weighting.
• line_thickness (1–5) - cosmetic; thicker lines read better on small screens.
• background (black/gray/white) - matching the background to your node theme.

There are no outputs - like Color Wheel View in the same pack, this is a terminal preview node. Wire an image in, look, and act. The productive loop: histogram → adjust with Levels Adjust / Curves Adjust Pro → re-check the histogram. Because it's preview-only, it never pollutes your batch outputs; the chart lives in the node body.

Install

It's part of the MachinePainting Nodes pack:

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

or ComfyUI Manager → "MachinePaintingNodes" → install → restart.

Gotchas

The one trap is confirmation bias: a "correct" histogram is not a universal goal. A moody low-key image is supposed to be left-heavy, and a bright airy look is supposed to be right-heavy. Read the histogram against your intent, not against an imagined ideal. Also, if the chart looks blank on a very dark image, that's the normalization compressing everything into a pixel-wide sliver - bump the line thickness or switch background to white for contrast. And remember it shows distribution, not spatial position - it tells you that highlights are clipped, not where. For "where," you want the image itself or a mask overlay.