The mix
Split into parts
When the voice should be its own fader.
Split into parts takes the bounce and makes stems: voice, drums, bass, and the rest. Do this before the finish button when you want the voice on its own fader. Then press Make it sound finished again. The original bounce is turned down so the parts are what you hear.
By default the split runs on this machine when the machine can keep up. Cleaner split on our server sends it to a neural model (Demucs) instead, which takes about a minute. Cancel stops a split that is still running. The thin bar under the transport is the progress.
The split on your computer is a fast filter: parts you can balance, not clean isolation. The server split is a neural model with much cleaner parts. Where splits run is the full story.