Singing along to what is already playing

How it goes

Play a song wherever you normally play songs. Press the button in karanow. About two seconds later the lead vocal is gone and the band is still going; you sing over that. The original singer can be blended back at any level with a slider, which is what you want when you are still learning the tune.

There is nothing to import and nothing to choose from. The song does not need to be in a catalogue, because there is no catalogue.

Why no app has to support it

karanow reads the Mac’s own output stream — the sound already on its way to your speakers — rather than talking to any application’s API. That is why Spotify, Apple Music, a browser tab, a video player and anything else that makes noise all work identically, and why none of them can break it by changing an interface.

Some applications publish what they are playing, and those get more: with Spotify and Apple Music, karanow can read the current track, follow the progress, and take pause, next and previous. Applications that publish nothing still get real-time accompaniment; they just do not get transport control.

About the two seconds

Separation works on a moving 1.5-second window, so the first output arrives about two seconds after you press the button, occasionally closer to three on a busy machine. After that it stays continuous and everything — the backing, the notes, your own pitch line — sits on the same delayed timeline, so nothing drifts against anything else.

Sound quality, honestly

Real-time separation is not as clean as an offline pass, because an offline tool can look at the whole song and take as long as it likes. On dense mixes you will hear traces of the vocal, and on sparse ones the result is close to transparent. Whether that matters depends on whether you are practising or recording; for practising, it stops mattering within about a verse.