(SP) Baxandall EQ

The shortest path from "generated audio that sounds like a laptop speaker" to "acceptable." (SP) Baxandall EQ is the two-band little brother of the pack's 3-band EQ: a low shelf and a high shelf, nothing else. No mid band, no frequency knobs, no Q. You get exactly two inputs that matter and one job - tilt the tone without thinking too hard about it.

The 3-band version exists and is genuinely more useful for real shaping, so why use this one? Because sometimes two knobs is the whole point. The low shelf is fixed at 100 Hz, the high shelf at 10 kHz (the code hard-codes both corner frequencies). That's the classic "make it warm, make it less harsh" sweet spot for a single vocal or a synth pad, and with the corners locked there's exactly one way to use it wrong: cranking both gains to +20 dB and wondering why it sounds like a tin can with a fever.

What you're actually setting

• bass_gain_db - boost or cut below the 100 Hz shelf, ±20 dB, default +1 dB. This is your "less mud in the low end" or "more sub weight" knob.
• treble_gain_db - the same for the 10 kHz shelf, ±20 dB, default +1 dB. Your "tame the fizz" or "add air" knob.

Everything feeds a single processed_audio output. That's the whole node.

Under the hood it's the standard RBJ Audio EQ Cookbook shelf equations (slope S=1.0, the gentle musical slope), applied as biquads via torchaudio. And like every other processor in this pack, it measures your input's LUFS loudness first and normalizes the output back to the same level afterward - so boosting bass doesn't secretly make the whole thing louder and fool your ears during A/B. That's a thoughtful default you'll appreciate the first time you bounce between +1 and -4 dB on the treble shelf and the perceived volume doesn't jump.

When to grab it vs. its sibling

Honest take: if you're already loading this pack, just use the (SP) Baxandall 3 Band EQ - it does everything this does plus a mid band. Reach for the 2-band when you want a blunt instrument that can't be over-tweaked: a quick low-pass-ish tilt on a bass line, taming a harsh TTS voice, or as a one-stop "make it less cheap" before a Convolution Reverb. Fewer knobs means fewer places to make it worse.

Installing it

ComfyUI Manager → "Install Custom Nodes" → search ComfyUI Signal Processing, or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/c0ffymachyne/ComfyUI_SignalProcessing

then restart ComfyUI. Dependencies come from the pack's requirements.txt (scipy, pyfar, torchaudio, pyloudnorm, cupy-cuda11x, etc.) - Manager handles them, or pip install -r requirements.txt manually. The repo labels itself a work in progress, and since the pack imports the CuPy-based Limiter at startup, a CUDA/cupy version mismatch is the most likely reason the whole pack (this node included) won't appear in your node list. If you hit that, fix CuPy first and the EQ shows up with everything else.