LC Audio EQ 3-Band 🎛️

Generated soundtracks have a sound problem in common: they're often boxy or dull, with everything sitting in the muddiest part of the spectrum. LC Audio Equalizer 3-Band is the corrective - a three-knob EQ that can add low-end body, cut the honky midrange, or put a little air on top without making you install a full DAW-grade audio stack into ComfyUI.

The three bands map to the three fixes people actually need. Low shelf to thicken or thin the bass. A peaking mid to tame "boxiness" (that's the tooltip's word, and it's the right one - voice and cheap speakers live around 800 Hz–1.5 kHz, and a small cut there cleans up most AI audio). A high shelf for sparkle above ~5 kHz or to dull hiss. The defaults - 100 Hz low corner, 1 kHz mid, 5 kHz high - are sensible starting points you'll tweak by ear.

How it works

This is real filtering, done in pure NumPy: low shelf, peaking band, and high shelf implemented as biquad filters, applied per channel for stereo, all in-memory with zero external dependencies (no SoX, no Numba). The behavior that separates a well-made EQ from a toy is the bypass: each band only runs when its gain isn't 0 dB. Leave all three at 0 and the node is a pass-through - critically, it doesn't "color" the signal with the shelf/peak math when you're not using a band. So you can wire it in permanently and dial bands in and out as needed.

Gain range is −24 to +24 dB per band, generous. mid_q controls the width of the mid peak: 0.707 is a broad, musical cut/boost, higher Q narrows it toward a surgical notch.

The inputs that matter

• low_gain_dB / mid_gain_dB / high_gain_dB - the three gains. 0 skips that band. Small values go a long way; +3 dB of shelf is a noticeable change.
• mid_freq / mid_q - center and width of the mid band. To de-box a voice, cut around 300–500 Hz slightly; presence lives ~2–4 kHz.
• low_freq / high_freq - the two shelf corners, if you want them somewhere other than the defaults.

One output: the processed audio.

How to install it

It's in the lonecatone23/ComfyUI_LC_AV_nodes pack (LC Audio_Video, MIT, by lonecatone23 - companion to the image-side ComfyUI_LC123_nodes, kept as a separate repo):

• ComfyUI Manager: search "LC Audio_Video" or ComfyUI_LC_AV_nodes.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/lonecatone23/ComfyUI_LC_AV_nodes, restart ComfyUI ([LC AV] total 16 nodes in the console).

No model downloads, no extra pip installs - just ComfyUI's built-in Python.

Where it's easy to go wrong

The number-one beginner move is cranking a band and wondering why everything distorts. Remember there's no clip protection on the EQ - boosts can push peaks past full scale, so pair it with LC Audio Volume (turn down after boosting, or keep boosts modest). And trust the "small cuts to tame boxiness" advice in the tooltip: subtractive EQ almost always sounds better than additive when you're cleaning up an AI mix. If you need a frequency analyzer to see what you're doing, this pack doesn't ship one - do your first pass by ear, since the spectrum problems in generated audio are usually obvious enough.