Audio Split 2 Channels

The inverse of Join 2 Channels is Split 2 Channels (SET_AudioSplit2Channels), and it's exactly what it sounds like: feed it stereo, get back two mono signals, one for the left channel and one for the right. It's the "unzip" step for audio, and it exists so you can treat the halves of a stereo recording as independent signals.

How it works

One input - audio - a stereo signal. Two outputs: audio_left (mono, the left channel) and audio_right (mono, the right channel). It's a straight channel extraction, no processing, no resampling, no downmixing. Same sample rate, same length, just split.

Here's the thing the README says flat out: the node raises an error if the input isn't 2-channel stereo. No downmixing of 5.1, no fake-stereo guessing, no averaging a 6-channel file into two vague channels. If you feed it mono, it refuses to run. That's a deliberate design choice - a stereo split is only meaningful when there are exactly two channels - but it means you need to know what's on the wire before you reach for it. If your input might be mono, run it through a Channel Converter (force_stereo) first so the split always has something to split.

When you'd use it

The classic pattern is split–process–rejoin, the stereo-audio equivalent of processing RGB channels separately. Split a stereo recording into left and right, apply different processing to each half (different gains via Blend, different cuts), then rebuild stereo with Join 2 Channels. You can also split to inspect: if one channel of a recording is broken or silent, splitting isolates it so you can diagnose it - or deliberately replace it.

One honest note: most stereo music isn't "left instrument / right instrument." It's a mix across both channels, so splitting and processing halves independently can sound odd compared to what you expect. Where it shines is with genuinely channel-separated content - dual-language recordings, binaural-ish captures, or material you generated as two distinct mono stems in the first place.

Install & gotchas

It ships in the Audio Batch pack. ComfyUI Manager search "Audio Batch", or:

cd ComfyUI/custom_nodes/
git clone https://github.com/set-soft/ComfyUI-AudioBatch
pip install seconohe

Restart ComfyUI. No model downloads.

The gotcha is the strict stereo requirement - the error is loud, which is the good kind of failure, but it means you'll want a converter upstream if your audio's channel count is ever in doubt. And remember that both outputs are mono: don't feed them into something that expects stereo and wonder why it sounds collapsed. If you're splitting just to rejoin unchanged, you're doing extra work for nothing - this node earns its place in a graph when each half goes somewhere different.