Audio Bass/Treble

Audio Bass/Treble is the smallest node in Audio General (niknah's audio-general-ComfyUI pack), and it's exactly as simple as it sounds: pick bass or treble, set a gain in dB, and get the same audio back with that frequency range boosted or cut. There's no EQ curve to draw, no multiband anything - it's a single shelf filter, on purpose. Reach for it when your TTS voiceover sounds thin or muddy and you want it to sit better against a music bed, or when you need to de-emphasize the hiss in a recording by cutting the treble shelf a few dB.

How it works

Under the hood it's a thin wrapper around torchaudio's bass_biquad and treble_biquad - second-order IIR filters. frequency_type picks which band you're shaping, gain sets how many dB that band gets boosted (positive) or cut (negative), and the filter adjusts the region around a center frequency. Defaults: bass is centered at 100 Hz, treble at 3000 Hz, and central_freq lets you override either. Q (default 0.707) is the filter's resonance - the width of the affected band, the same Q factor you'll find on any synthesizer filter.

The inputs that matter

• gain - the dB of boost or cut, from -50 to up. The trap: the default is 15 dB. That is a lot - it's the "party mode" preset. Start around 3–6 dB and you'll get a tasteful warmth or presence; 15 dB of treble will make a voiceover genuinely harsh. This is the first thing to change after you wire it in.
• frequency_type - "Bass" or "Treble". That's the whole choice.
• central_freq - leave at -1 to use the defaults unless you specifically want to move the band (e.g. push the bass shelf down to 60 Hz for a sub-heavy music bed).
• Q - leave it. 0.707 is the classic shelf response; you'll know if you need to touch it.

Output is a single AUDIO, same sample rate, same length - drop it anywhere in the chain between your audio source and PreviewAudio/SaveAudio.

Install

The pack installs wholesale: ComfyUI Manager → search "Audio General" → install, or

cd ComfyUI/custom_nodes
git clone https://github.com/niknah/audio-general-ComfyUI

Restart ComfyUI. Dependencies are torchaudio, librosa, torch-time-stretch, audiostretchy - no models, CPU-only.

Where people get burned

Almost entirely by that 15 dB default - the node's out-of-the-box sound is way too hot, so budget a few seconds to pull gain down before you judge it. And don't expect a full parametric EQ; you get one band, no high-pass, no low-pass, no mid. If you need more control, chain a couple of these or reach for a proper EQ node. For "make this voiceover less boring" though, a couple of dB of bass or treble through this node is the fastest fix in the pack.