CombineAudioNode

CombineAudioNode is the quiet other half of the VocalSeparation-ComfyUI pack. Its sibling splits a track into vocals and instrumental; this node takes those two stems - or any two audio signals - and adds them back into one track. That's the whole job, and it's easy to underestimate how often you need exactly that.

What it actually does

Two AUDIO inputs (vocal, instrumental), one AUDIO output. It resamples both to 44.1 kHz, pads whichever is shorter with silence so they line up, then sums the samples. No AI, no model, no downloads - it's a two-input mixer wearing a node costume. Under the hood it's just resample-and-add.

Why you'd reach for it

The obvious use is inverting a separation. Split a track with VocalSeparationNode, process the vocal (pitch-shift it, slap on reverb, clean up a take), then combine the treated vocal with the untouched instrumental and walk away with a "new mix" without leaving the graph. That split → process → recombine loop is the whole reason this pack ships a combine node.

It's also a dirty trick enabler. Take an acapella you like, split out its instrumental, and combine the acapella with a different instrumental for a DIY re-voice. And because it accepts any AUDIO, not just its sibling's output, it's a clean way to sum two audio signals from anywhere in your workflow onto a single wire.

The catches

All real, all worth knowing before you build a workflow around it. First, it's a straight sum - no leveling, no EQ, no fade. If the two inputs sit at wildly different loudness, you get a louder, possibly clipping result. Second, the splitter outputs mono stems at 44.1 kHz, so the combined result is mono 44.1 kHz; this is a utility node, not a mastering chain. Third, mismatched lengths get the shorter one zero-padded and the result runs as long as the longer input - if the tracks aren't the same duration, you'll hear dead air where the pad is.

Install

One pack covers both nodes, so install it once:

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

Or skip all that and search "VocalSeparation-ComfyUI" in ComfyUI Manager, then restart. The combine node itself is pure numpy plus torchaudio, which you already have - but here's the rub: you can't install it alone. The pack's one requirements.txt drags in demucs, asteroid, transformers, and friends even if you never run the splitter. That heavy install is the price of the pack, not of this node.

It's the most boring node in the pack, and the one that makes the splitter genuinely useful.