Merge Audio

AudioMerge is the node for the moment you realize a track needs to sit under another one rather than after it - a music bed under a voiceover, a whoosh layered over a scene cut, a synth pad behind a generated melody. It takes two AUDIO inputs and overlays their waveforms into one, and of the whole audio editing family it's the one with real creative weight.

The description in the node is exactly right: "Combine two audio tracks by overlaying their waveforms." It's the audio analogue of an image composite. You feed it two tracks, pick how you want them combined, and get back a single mixed track that's the same length as the longer of the two inputs.

How it works

The mixer does four steps, and knowing them saves you from a couple of surprises. First it resamples both tracks to the higher sample rate if they differ - 44.1 kHz and 48 kHz sources merge cleanly with no click or pitch shift. Then it fits audio2 to audio1's length: if audio2 is longer it gets trimmed, if it's shorter it's zero-padded to fill. Then the actual combination happens, using the method you picked. Finally, if the combined waveform peaks above 1.0, the whole thing gets normalized down so you don't clip on export.

The merge_method combo is where the character is:

• add - straight sum. The default vibe; two tracks playing on top of each other.
• mean - the average of the two. Louder signals aren't pushed to clipping as fast, so this is gentler and often the better default for voice-over-music.
• subtract - track 1 minus track 2. Niche, but genuinely useful for removing a known noise profile or phase-canceling a hum.
• multiply - one track gates the other. More an audio-geek toy than a daily tool, but fun for tremolo-style effects.

Inputs and output

• audio1 (AUDIO) - the base track. Its length sets the output length.
• audio2 (AUDIO) - the track layered on top; trimmed or padded to fit audio1.
• merge_method (COMBO) - add, mean, subtract, or multiply.
• AUDIO (output) - the mixed result, ready for PreviewAudio, a save node, or further processing.

Getting it

Ships with ComfyUI core; it's part of the batch of audio editing nodes that landed in September 2025. Update to any recent core and it's there. Nothing to install.

Common issues

The length behavior trips people up more than anything else. AudioMerge doesn't produce a track the length of both inputs put together - that's AudioConcat's job. It produces a track the length of audio1. If your mix comes out shorter than you expected, that's why. And if you want the two tracks to be balanced, level them first with AudioAdjustVolume; a loud music bed will happily bury a voiceover and leave the add result pumping against the normalization ceiling.