Audio Gain + Pan

Volume and pan are the two things you adjust on basically every piece of audio, and up until recently there was no good reason to do either inside ComfyUI - then video workflows started carrying soundtracks and dialogue, and suddenly you want a voice quieter than the music without leaving the graph. MKRAudioGainPan is that: one node, a gain knob and a pan knob, plus a couple of stereo conveniences. It's the kind of boring utility that quietly does exactly what it says.

It lives in the Utility branch of the MKRShift_Nodes audio set, so it's the stage before FX rather than the FX itself. Typical use: pull dialogue down with gain_db, push a voice a touch off-center with pan, then feed the result into MKRAudioMix to combine it with a bed track.

How it works

Gain is straightforward dB scaling (0 dB = unchanged, −6 dB halves the level, +6 dB doubles it). Pan uses a constant-power law - the standard studio trick where the left and right channels get cos(angle) and sin(angle) gains so the perceived loudness stays constant as you move across the field - rather than a naive linear fade that makes the middle sound louder. pan runs −1 (hard left) to +1 (hard right), 0 dead center.

Two stereo conveniences make it more than a pair of knobs:

• force_stereo - turns a mono input into a real two-channel signal (necessary before panning a mono voice, and for getting a mono track into a stereo mix).
• normalize_peak - rescales the result to just under full scale. If you've boosted by a lot and don't want to clip, flip this on.

Inputs that matter

• audio - MKR_AUDIO payload, waveform tensor, or file path.
• gain_db - −60 to +24 dB. The volume knob.
• pan - −1 to +1. The position knob.
• force_stereo / normalize_peak - the two toggles above.
• Standard save block: output_format, filename_prefix, subfolder, overwrite, optional filename_label.

Outputs: audio (MKR_AUDIO), output_path, duration_sec, summary.

Install

ComfyUI Manager (search "MKRShift Nodes"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/criskb/MKRShift_Nodes

Restart. Pure numpy math - no ffmpeg required for the processing itself; it only matters for compressed output formats. No models.

Common issues

The trap is panning a mono source and hearing nothing move: with force_stereo off, a mono waveform just gets a shared gain and pan does effectively nothing. Turn force_stereo on first. Second, watch the cumulative math - if you've boosted +6 dB and your upstream was already hot, normalize_peak is off by default and you will clip; the summary output shows you the peak if you inspect it. And negative pan means left, positive means right - the old DAW convention where "pan left" was negative catches people constantly.