FUDA Amplitude Visualiser

The FUDA Amplitude Visualiser is the debugging companion to the FUDA pack's color-transfer node. It renders the centered Fourier amplitude spectrum of any image as a grayscale heatmap - bright means high energy - so you can see the frequency-domain signature that the FUDA Image Adaptation node operates on. It doesn't edit anything. It's a window into what the math is doing.

Why it's useful

The whole premise of the FUDA nodes is that an image's amplitude spectrum carries its color and lighting style, and the adaptation node mixes the low-frequency part of one image's spectrum into another. That claim is a lot easier to trust when you can look at it. The pattern the README suggests: wire your source image into this node and preview it, then wire the adapted output through the same visualiser - you should see the bright low-frequency center of the spectrum shift toward the reference's signature. If it doesn't, that's your first clue something about the transfer isn't what you expected.

It's genuinely handy for comparing domains too - two images with matching subject matter but very different lighting will show clearly different amplitude patterns, which is the sort of thing that's hard to articulate but easy to point at.

How it works

The node FFTs the image, shifts the DC component to the center, takes the magnitude, and averages across color channels to get a single greyscale spectrum. If log_scale is on (default), it applies log(1 + amplitude) - without that, the huge DC peak at the center blows out everything else and the spectrum is just a white dot in a black field. The result is min-max normalized per image and returned as a proper 3-channel IMAGE.

The only inputs are image (any IMAGE) and the log_scale boolean toggle; the only output is amplitude_map, which wires straight into a Preview or Save Image node. Turn log_scale off once to see why it defaults to on - the difference is the entire reason the toggle exists.

Install

It ships in the same pack as the adaptation node, so the install is identical:

cd ComfyUI/custom_nodes
git clone https://github.com/bemoregt/ComfyUI_FUDA

or find FUDA in ComfyUI Manager, then restart. No extra dependencies, no model files - torch and numpy only. All three nodes appear under the FUDA category.

Gotchas

This is a pure visual utility, so there's not much to trip over. One honest note: the spectrum is per-image normalized, which means you're comparing shapes, not absolute energy levels - a dim image and a bright image both fill the range. If you're comparing before/after, keep the same image pair and just accept that the heatmap shows relative structure. And as with the rest of this young, single-commit pack, there's no community track record to lean on yet - but for a node this simple, that's a small risk.