Cut Audio Cuts JK🐉

The inverse of the other JK audio cutters: instead of pulling out one segment, Cut Audio Cuts JK takes the entire SCENECUTS list and concatenates every cut back together in chronological order into a single audio file. At first glance that sounds like re-arranging deck chairs - isn't the full audio already the full audio? Not once you've filtered cuts, it isn't.

This is the node you use when you've made your scene list, dropped some cuts (via select_cuts on Scene Cuts JK, or by generating only some segments), and you want a single "what the final video will actually hear" track. The cuts are sorted by start time, each one is sliced out, and they're glued end-to-end with torch.cat. Whatever you didn't select just disappears from the timeline. Think of it as the audio counterpart of a video editor's "remove unselected clips and close gaps."

There's one detail that matters when your scene cuts come from the Wan video world: warmup handling. The node adds the warmup offset from the SCENECUTS metadata to each cut's start time, shaving off the warmup frames Wan Wrapper sometimes needs before normal output. So the stitched audio actually lines up with the frames that survive to the final video - a subtle bit of accounting that stops you chasing a sync drift that isn't yours.

Inputs are minimal - just audio and scene_cuts, one audio output. There's genuinely nothing to configure, which is refreshing: either you need the merged track or you don't.

Install. It's part of ComfyUI-JakeUpgrade:

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

Or via ComfyUI Manager → "JakeUpgrade". Windows portable: install.bat or ../../../python_embeded/python.exe -s -m pip install -r requirements.txt. Restart. Light dependencies, no models.

Gotchas: it raises an error if scene_cuts has zero cuts - empty selections are a failure, not a silent empty track, so filter your cuts upstream carefully. And remember the SCENECUTS type only exists if the pack's video module is loaded; if you restricted modules in config.ini, add video back. The usual pack notes apply: LOAD_DEPRECATED_NODES = True for missing-JK-node reports, and disable ComfyUI MultiGPU per the README if you run into CUDA errors on current ComfyUI.