Audio Split

Audio Split is the simplest node in the ComfyTV audio suite, and it knows it. It cuts a track at one timestamp - split_s in seconds - and gives you back two halves: audio_a (everything before the cut) and audio_b (everything after). That's the whole spec.

Don't mistake the simplicity for uselessness. This is the workhorse behind "I like the intro of this track but want to loop just the drop," or "let me process the verse and the chorus differently before I mix them back together." Two of these in sequence turn one track into four regions, each independently processable - and because ComfyTV stages re-run on their own (the per-node Run model, not the global queue), you can tweak the split point or the processing on one half without re-rendering everything above it.

Input is audio or video (it splits the video's audio), and the three hidden inputs (force_run_token, project_id, parent_output_id) are ComfyTV's internal plumbing - leave them alone. The output halves are the same type you put in, so they feed straight into Audio Mix, Audio Repair, an FX Chain, or a Video Stage's audio input.

The one thing to remember: split_s is an absolute timestamp, not a percentage. A 60-second track at 30 means a clean halfway cut; a 3-minute track at 30 cuts ten seconds in. If you're splitting at the same point across many tracks, fine - but if you want proportional splits, do the math first. The Audio Split Export stage is the sibling for when you want silence detection and file export instead of a single manual marker; this node is for the fast, deliberate, in-pipeline cut.

Install is the pack-wide story: ComfyUI Manager search "ComfyTV", or git clone https://github.com/jtydhr88/ComfyTV into custom_nodes/ and fully restart the backend. On Desktop/macOS with multiple ComfyUI installs, clone by absolute path into the running instance or the pack silently won't load. Zero extra pip dependencies - it's a numpy trim in the DSP layer, no models, CPU-only.