(SP) Load Audio

Every pipeline in this pack starts here. (SP) Load Audio is the front door: it pulls a file off your disk and hands it to the rest of the Signal Processing nodes as a proper ComfyUI AUDIO object. No AI, no model download, no API key - it's torchaudio.load() with a nicer wrapper.

The catch, and it's the one thing that bites everyone the first time: this node doesn't read audio from anywhere on your machine. It only sees files inside a single folder: ComfyUI/input/samples/. Drop your .wav, .mp3, .ogg, .m4a, .flac, or .mp4 in there, refresh the node, and it shows up in the dropdown. That's the whole trick - a surprising number of first-run "why is my file not in the list?" threads are just people pointing at their desktop instead.

What you set

• audio_file - the dropdown of files it found in input/samples. Pick yours.
• gain - a simple linear volume multiplier, 0.0–8.0, default 1.0. Use this for a rough trim before normalization; it's a plain multiplication, not a fancy loudness control.

One audio output comes out the right side, ready to plug into (SP) Filter, PaulStretch, the EQs, or anything else in the pack.

Two things worth knowing

The README's description for this node is flat-out wrong - it says this node "lets you stretch audio to about 100x its original speed while maintaining pitch." That's PaulStretch, not Load Audio. The description got copy-pasted into the wrong section, and the README also mixes up the gain field ("when to start audio from" is nonsense). Ignore all of it; the behavior above is what the code actually does.

Also: loading a mono file gives you a mono AUDIO. Most of the pack's processors work fine with mono (a few, like Stereo Width, will happily convert mono to stereo for you), so don't panic if your source is a single channel.

Installing it

Same as the rest of the pack. 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 requirements.txt (scipy, pyfar, numpy, torch, torchaudio, pyloudnorm, cupy-cuda11x); Manager handles them, or pip install -r requirements.txt manually. This is an early WIP repo, and if the whole pack won't load, a CuPy/CUDA version mismatch from the pinned cupy-cuda11x is the usual suspect. Create ComfyUI/input/samples/ if it doesn't exist, drop your file in, and you're off.