Load Audio Plus From Path (UTK)

ComfyUI's built-in audio loader is a file-picker widget, which is fine until you want to script a batch, point at a path from a text node, or process the same file repeatedly. Load Audio Plus From Path (UTK) takes a plain filesystem path and loads audio from it - and does the preprocessing on the way in. Gain, offset, duration, resample, stereo conversion. It's the "plus" in the name, and it's why this is the audio input node from ComfyUI-UniversalToolkit I'd default to for any real audio pipeline.

The pack's audio tree is one of the less common ones in the ComfyUI world, so this node quietly covers a gap most people don't realize exists until they need it.

How it works

It uses librosa for the heavy lifting and soundfile for the file I/O, loading whatever you point at (mp3, wav, flac, and friends). The path is cleaned up first - quotes stripped, backslashes normalized to forward slashes, so a Windows path pasted from Explorer works without surgery. Then it applies the processing chain in order:

• offset_seconds - skip this much in.
• duration_seconds - only keep this much (0 means "the whole file," which is the convention to remember).
• gain_db - volume in decibels, converted via the standard 10^(db/20) math. −100 to +100, 0 = unchanged.
• resample_to_hz - change the sample rate; 0 means leave it alone.
• make_stereo - force two channels if the source is mono.

Then it returns the audio in the standard ComfyUI AUDIO format plus the sample_rate, channels, and duration as plain numbers, which are handy for downstream scheduling or display.

One smart detail: the node keys its IS_CHANGED on the file's modification time. Touch the file on disk and ComfyUI automatically re-runs the pipeline without you having to mute/unmute the node to force it.

Inputs and outputs

• path (STRING, default ./audio.mp3) - the file path.
• gain_db, offset_seconds, duration_seconds, resample_to_hz, make_stereo - the processing chain.

Outputs: audio (AUDIO), sample_rate (INT), channels (INT), duration (FLOAT).

Install and gotchas

Install via ComfyUI Manager (search "ComfyUI-UniversalToolkit") or:

cd ComfyUI/custom_nodes
git clone https://github.com/whmc76/ComfyUI-UniversalToolkit
pip install -r requirements.txt

Restart ComfyUI. No models. This node does need the audio dependencies - librosa, torchaudio, soundfile are all in the pack's requirements.txt, and the pack's loader wraps the audio module in a try/except, so if librosa is missing you won't get a loud error, just no audio nodes. That silent-missing-node failure is the classic way this pack bites; if audio nodes are absent after install, pip install -r requirements.txt didn't complete.

Other gotchas: duration_seconds: 0 means "everything," which is counterintuitive - set it to 0 and you get the full file. And path errors are the number-one support question: the node validates the file exists and gives you a detailed message, but it can't read your mind about absolute vs relative paths. ./audio.mp3 resolves relative to where ComfyUI was launched, so an absolute path is the safer habit.