Split Audio

SplitAudio cuts one AUDIO track into two at a timestamp you pick. It's the inverse of the join nodes in this pack - where ConcatAudio and JoinAudio put clips together, this one takes a clip apart so you can process the halves differently, drop the boring middle, or feed each piece into its own branch of a workflow. It ships in lebrosoft/ComfyUI-VideoChatWrapper, a small pack whose audio utilities are a straightforward torchaudio toolkit hiding behind a video-chat name.

How it works

The mechanism is about as direct as it gets: pick a time in seconds, convert it to a sample index, and slice the waveform tensor. Everything before the cut goes to audio1, everything after goes to audio2. Both outputs keep the original sample rate, so they stay compatible with every other audio node in the pack.

One thing to know before you rely on it: this is a hard cut. There's no fade, no crossfade, no smoothing at the splice point. If you're splitting a clip where a hard edit will pop or click, you'll want to add fades with a different pack. For cutting out a segment or isolating the first half of a file, it's exactly right.

The inputs

• audio - the track to split.
• second - the split point in seconds, a float. This is the one number you set, and it's the entire decision.

Two outputs, both AUDIO: audio1 is the part before the cut, audio2 is the part after.

A trap worth flagging: the default for second is 0, which splits at the very start - you get an empty first half and the whole track as the second. If you leave it at the default and only look at audio2, you'll think the node did nothing. Set the number.

Why you'd reach for it

• You have one long recording and want the speech separated from the silence, so you split and analyze each half.
• You're building a workflow that processes the intro differently from the body.
• You want to excise the first N seconds of a generated clip.

In each case the two outputs slot straight back into the pack's other audio nodes, or into any node that consumes AUDIO.

Installing it

Standard for this pack. ComfyUI Manager → search ComfyUI-VideoChatWrapper → install, or:

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

Restart ComfyUI. The audio half needs torchaudio, bundled with ComfyUI's portable install (and pip install torchaudio if your own environment lacks it). The requirements.txt only lists qwen_vl_utils for the video-chat nodes, so audio users pull no extra dependencies and no model weights.

Worth saying plainly: this pack has no community footprint to speak of - zero usage stats, no tutorials - so the troubleshooting support is "you and the source code." The upside is the source code is short and readable, and for a hard-cut split there's genuinely not much to get wrong.