Load Audio File

LoadAudioStandalone is the pack's "Load Audio File" node, and it does exactly one useful thing: turn a file on disk into the AUDIO format every other node in this pack expects. You'll mostly reach for it when you want to feed a reference track into HuggingFaceMusicGen's conditioning_audio input for continuation, or when you want to run existing audio through the loop-processing and preview nodes without generating anything new.

The one thing to know before you use it: there's no file browser. The single required input is audio_path, a plain STRING. You type (or paste) the absolute path to your file, and the node loads it. No dropdown, no drag-and-drop, no "input/" folder scanning - that's the design, and it's the main thing that trips people up. If you want the ComfyUI file-picker experience, the core LoadAudio node exists for that; this one is for when you already know the path, which is precisely what the conditioning-queue system needs when it loads files programmatically.

How it works

Under the hood it decodes with PyAV (av), pulls the first audio stream, converts to float32 PCM, and hands you back a standard ComfyUI AUDIO dict: {"waveform": ..., "sample_rate": ...}. It handles whatever sample rate the source file has - the resampling to MusicGen's 32kHz happens later, in the generation node, not here. And because it's PyAV, "audio file" means more than wav: mp3, flac, opus, m4a, whatever PyAV can decode. It raises a clear error if the path doesn't exist (Audio file not found: <path>), so a dead path fails loudly rather than silently feeding you silence.

Input: audio_path (STRING). Output: AUDIO.

Where it fits

The typical wiring looks like this: LoadAudioStandalone → HuggingFaceMusicGen.conditioning_audio, then let MusicGen continue the track. Or LoadAudioStandalone → LoopingAudioPreview if you just want to loop and audition an existing clip before deciding whether to regenerate it. It's also the natural companion to the pack's own savers: save with SaveAudioStandalone, load the filepath back in elsewhere, and you've got a poor man's audio caching loop.

Install

Part of the ebrinz/ComfyUI-MusicGen-HF pack, so install once and you have all ten nodes. ComfyUI Manager → search "ComfyUI-MusicGen-HF", or:

cd ComfyUI/custom_nodes
git clone https://github.com/ebrinz/ComfyUI-MusicGen-HF
cd ComfyUI-MusicGen-HF
pip install -r requirements.txt

Restart ComfyUI after. The dependency that matters here is PyAV (av in requirements.txt).

Troubleshooting

Two failure modes dominate. First: file not found - you pasted a relative path or got the filename wrong. Give it the full absolute path, and on Windows remember backslashes need escaping. Second: decoding errors on weird containers - PyAV is finicky about some obscure codecs, but for standard wav/mp3/flac it just works. One more subtle thing: the output is mono-safe but will be whatever channel count the file has; if your file is stereo, downstream nodes that assume mono (like MusicGen's conditioning path, which takes the first channel) will quietly downmix for you. Nothing to fix, just know it's happening.