Concat Audio

ConcatAudio is the node you reach for when you've got two audio clips and you want them to play back-to-back as one track - a dialogue scene followed by a reaction shot, an intro stinger glued onto a narration, two takes you want in one file. It's part of the audio toolkit that ships inside lebrosoft/ComfyUI-VideoChatWrapper, a small pack that is really two packs in a trench coat: a video-chat pipeline and a set of generic torchaudio editing utilities. This node is from the second half, and it does one thing cleanly.

How it works

Under the hood it's dead simple. ComfyUI's AUDIO type is a dict with a waveform tensor and a sample_rate, and ConcatAudio just concatenates two waveforms along the time axis. If you set silent_interval to, say, 0.5, it inserts half a second of zeros between the two clips before joining them. There's no crossfade, no fade-in/out, no loudness matching - it's a hard splice with an optional gap of silence.

The inputs that matter

Only three, and one of them is the whole show:

• audio1 and audio2 - the two AUDIO tracks to join. The order matters: track 1 plays, then track 2.
• silent_interval - seconds of silence inserted between them, as a float. Default is 0. Set it to 0.25 or 0.5 if the cut feels cramped.

The single output is audio, an AUDIO track you can feed into whatever you use to save or preview audio in ComfyUI (this pack doesn't ship its own audio saver, so you'll pull that node from elsewhere).

Where people get burned

The error. If your two clips don't have the same sample rate - extremely common when one is AI-generated and the other came from a phone - the node throws sample_rate is different. Same story if one clip is stereo and the other is mono: audio1 is stereo but audio2 is monaural. It won't silently mangle your audio; it refuses and tells you exactly why.

The fix is to normalize both inputs before they reach this node. ResampleAudio (also in this pack) aligns sample rates, and ConvertAudioChannels handles the mono/stereo mismatch. That's the real shape of the workflow: two clips, a resample and a channel convert each, then ConcatAudio. Annoying the first time, automatic the third.

Installing it

The pack installs like any other custom node. Easiest is ComfyUI Manager - search for ComfyUI-VideoChatWrapper and hit install. Or do it by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/lebrosoft/ComfyUI-VideoChatWrapper

Then restart ComfyUI. The audio nodes need torchaudio, which ships with ComfyUI's portable install; if you're running a custom Python environment you may need pip install torchaudio yourself. The pack's requirements.txt only lists qwen_vl_utils (that's for the video-chat half), so audio-only users don't need to install anything extra.

One honest caveat: this is a tiny, low-traffic pack. The node has effectively zero community footprint, so don't expect much troubleshooting help online. But the code is short, readable, and stable, and for a simple splice it does the job without drama.