Plot Filter Bank

A filter bank is just a matrix of numbers. Plot Filter Bank renders that matrix as an image so you can actually see what you built. It's the smallest node in this pack - one required input, one IMAGE output, three cosmetic labels - and it's exactly the kind of tool you don't appreciate until you've built a mel bank and wondered why the top of your spectrogram looks the way it does.

How it works

It's a single imshow of the filter-bank weight matrix with the frequency axis oriented the right way up (low frequencies at the bottom, matching how the rest of the pack draws spectrograms). Each triangle in the bank shows up as a diagonal stripe of overlapping weights. The three optional inputs are pure cosmetics - title, xlabel, and ylabel, which default to "Filter Bank", "filter bin", and "frequency bin" respectively. No show_grid, no channel handling, nothing clever. The output is a standard ComfyUI IMAGE you can preview or save.

Why you'd reach for it

Two real reasons. First, debugging: if your mel spectrogram comes out with empty top bands or the filters clearly don't span the range you asked for, plotting the bank shows you instantly whether the problem is your f_min/f_max/n_mels choices or something downstream. Second, teaching yourself: the visual contrast between a linear bank (uniform diagonal stripes) and a mel bank (stripes crammed at the bottom, sparse at the top) is the fastest way to internalize what the mel scale actually does. It's also the easiest way to confirm the Nyquist gotcha - set f_max above the sample rate and you can watch the top filters get clamped.

One thing it won't tell you: it plots the weights, not the effect. If you want to see what your audio looks like after filtering, wire the filtered SPECT into Plot Spectrogram instead. The two plot nodes are complements - this one shows the lens, that one shows the view through it.

Installing

Same as every node in the pack - ComfyUI Manager search "comfyui-audio-processing", or:

cd ComfyUI/custom_nodes
git clone https://github.com/rhdunn/comfyui-audio-processing

Restart and it's there. It runs on matplotlib, which ComfyUI already ships, so no extra dependencies and no model downloads. This is a genuinely thin node - which is fine, because it's meant to be a quick look, not a feature. Two inputs to set, one image out, done.