Load Audio

LoadAudio is where every audio pipeline in ComfyUI starts, and it's the one audio node people actually search for. Given how invisible this corner of ComfyUI is, that makes sense: you can't audition a generated track, mix a voiceover, or feed audio to a video model without first getting a file onto the graph, and this is the node that does it.

The audio stuff is a relatively young part of ComfyUI core. It landed in mid-2024, alongside support for Stable Audio Open, when audio became a first-class citizen instead of something you hacked around as a video side-channel. LoadAudio is the front of that family, and the rest of the built-in audio nodes (TrimAudioDuration, AudioConcat, AudioMerge, the save and preview nodes) all plug into whatever it hands you.

How it works

Under the hood LoadAudio uses PyAV to open the file, grabs the first audio stream, and decodes it into a float32 PCM waveform tensor. It handles the common integer formats (int16, int32) on the way in, converting them to the floating-point range [-1, 1] that the rest of the graph expects.

The output is the standard ComfyUI AUDIO type: a dict holding a waveform tensor shaped [batch, channels, samples] plus the sample_rate. That AUDIO socket is the coin of the realm in this category - every audio node in ComfyUI core consumes or produces it, so LoadAudio's output drops straight into a save node, a mixer, a trimmer, or a PreviewAudio.

Here's a freebie the docs rarely spell out: the file dropdown isn't limited to audio. LoadAudio also lists video files from your input folder and pulls the first audio track out of them. Need the soundtrack of a clip as a mixing source? Point LoadAudio at the video and you're done, no ffmpeg gymnastics.

Inputs and outputs

• audio (combo) - the only input. A dropdown of files in your input/ folder, with an upload button for dragging in new ones. It filters to audio and video content types.
• AUDIO (output) - the decoded track, ready to wire anywhere.

Getting it

It ships with ComfyUI core - nothing to install. It's been there since the June 2024 audio rollout, so if your ComfyUI is from the last couple of years you already have it.

Common issues

The classic gotcha: LoadAudio's dropdown suddenly "has no button to select a file" when you open a workflow someone else shared. This is a real, recurring complaint. The button lives in the node's widget state, which doesn't always survive a workflow JSON round-trip. The fix that actually works is to right-click the node and pick Recreate (or choose the LoadAudio node from the menu again) so the widget rebuilds.

Otherwise the failure modes are boring and obvious: the file isn't in input/ (drop it there or use the upload button), or the file genuinely has no audio stream, which throws a "No audio stream found" error. And don't worry about sample rates - if you later merge this with a 48 kHz clip, the merge nodes resample automatically to the higher rate.