Removing vocals on a Mac
Two quite different things get called the same thing, and picking the wrong one wastes an afternoon. Start here: do you want a file, or do you want to sing?
If you want a file, do not use karanow
Wanting an instrumental you can keep — to put on a phone, hand to a band, load into an editor — means you want an offline separator. Ultimate Vocal Remover (UVR5) is free, open source and runs on a Mac; lalal.ai and similar services do it in a browser. They read a whole file, take as long as they like over it, and hand you a stem.
They will also do a cleaner job than anything real-time can. Given minutes instead of milliseconds, a separator can look at the whole track at once. karanow cannot: it has 1.125 seconds to separate 1.125 seconds of audio, or the sound breaks up. That constraint costs quality and buys something else.
karanow exports nothing. There is no file at the end. If that is what you came for, the tools above are the right answer and this page is not trying to talk you out of them.
What real time buys instead
You do not find the song first. You play it — in Apple Music, in Spotify, in a browser tab, in a video player — and press one button, and the vocal comes out of what is already going. Nothing is imported, nothing is exported, nothing is downloaded, and there is no catalogue that has to have licensed the track for you to sing it.
That last part is the whole point. A karaoke app can only offer songs its catalogue has cleared one at a time, which is why the song you actually want is so often missing. Whatever your Mac can play, you can sing to.
It also means the original vocal is still there if you want it. The same separation pass produces both stems, so a slider blends the singer back in at any level — useful when you are learning a line rather than performing it.
The honest comparison
Offline separators win on quality and give you a file. karanow wins on there being nothing to prepare, works with anything that makes sound, and shows you the notes while you sing. They are not competitors so much as answers to different questions.
If you find yourself wanting both — a file for later and something to practise against now — using both is entirely reasonable and neither one gets worse for it.
What karanow needs
Apple Silicon and macOS 14.2 or later. Intel Macs cannot run it: the separation model needs more than they have, and rather than degrade to old-fashioned mid/side cancellation and call it the same thing, it declines.
What happens to the audio while it runs is written out in full on its own page.