Split Audio Channels

SplitAudioChannels takes a stereo AUDIO tensor and separates it into its left and right channels, each as its own mono track. It's the "stereo to mono" node - the search alias ComfyUI gives it is exactly right - and it exists for the workflows where the two sides of a stereo recording are actually different signals that need to be processed independently.

When does that happen? More often than you'd think. A stereo recording where the left channel carries the voice and the right carries a different voice (two-mic setups, phone recordings, some voiceover captures). A stereo mix where you want to apply EQ to one side only. Or a downmix situation where you want to drop one channel entirely. The node's job is to hand you the two channels so you can do that, and its sibling JoinAudioChannels does the reverse - which makes the pair a complete round-trip: split, process each side, join back into stereo.

How it works

Mechanically it's trivial: the node checks the input is stereo, slices the waveform's channel dimension into two [batch, 1, samples] tensors, and returns them as two AUDIO outputs. Sample rate is preserved on both.

The one hard requirement, and it's a real one: the input must be stereo. Because it's slicing two channels apart, feeding it a mono track throws an error - "Input audio must be stereo (2 channels)". If you're not sure what you've got, and the source is a video or a mixed file, that's an easy thing to trip over.

Inputs and output

• audio (AUDIO) - the stereo track to split.
• left (AUDIO) - the left channel, as mono.
• right (AUDIO) - the right channel, as mono.

Note the outputs have lowercase names (left/right), which is just ComfyUI's display convention for this node - the sockets are still standard AUDIO.

The workflow shape to keep in mind: split → process each side with whatever you like (EQ, volume, even a different trim per side) → JoinAudioChannels back into stereo. Since both split outputs preserve the original sample rate and sample count, the join is lossless - no resampling, no length trimming - as long as you don't change duration mid-path.

Getting it

Ships with ComfyUI core; part of the September 2025 audio editing batch. No install.

Common issues

The stereo-only constraint is the whole troubleshooting story. If you hit the channel-count error, your source isn't stereo - check the file or route it through a stereo source. One practical thing to know: because each output is mono, you can't feed either straight back into SplitAudioChannels, and if you feed a split channel into AudioConcat or AudioMerge it'll be duplicated to stereo automatically (the processing nodes assume stereo internally). That's fine - it means a split-then-process-then-join round trip through JoinAudioChannels works cleanly as long as you don't change the channel count mid-path.