Skip to main content
Audio Stem Split

Audio Stem Split · screenshot coming soon

​
What this node does

Audio Stem Split runs source separation on a mixed track and hands back individual stems. It uses an HDemucs model to pull apart the mix into vocals, drums, bass, and “other”, plus an accompaniment stem (the full mix minus vocals). You can then process or re-mix each part on its own. It accepts a COMFYTV_AUDIO snapshot or a COMFYTV_VIDEO (its audio track is used; audio takes priority). It outputs five COMFYTV_AUDIO stems and has a ▶ Run — separation is a neural model, so the first run downloads weights and each run takes real compute time. To feed native ComfyUI AUDIO in or out, insert a Bridge — see bridges.md.

​
When to use it

  • Pull an isolated vocal out of a song for a remix or acapella.
  • Remove vocals to make a karaoke/instrumental version (accompaniment).
  • Re-balance drums or bass independently of the rest of the mix.

​
Parameters

This node has no adjustable parameters — wire in a source and run. The separation model does the rest.

​
Outputs

​
Tips

  • Separation is never perfect — expect some bleed between stems, especially on dense or heavily-processed mixes.
  • The first run downloads the model; later runs are faster but still compute-heavy. Longer tracks take proportionally longer.
  • Feed the stems back into Audio Mix to rebuild a mix with your changes.
  • Audio Mix — recombine the stems (with new levels/pan) into a fresh mix.
  • Noise Reduction (Spectral) — clean up a stem after separation.
  • Audio Beats & Notes — analyze the isolated drum stem for tighter beat detection.