I'm afraid that the chances of being able to do this are not good. In fact they are awful. In theory if you have two versions of a track, one with a vocal and one without, then you can subtract one from the other, and be left with the vocal. But there are some rather stringent conditions that this trick has to meet in order to work. The first is that the music has to be absolutely identical - IOW the same track (this is pretty rare with karaoke tracks), it has to be the same amplitude - exactly - and lined up to be sample-accurate. Without all of this happening it will fail - miserably.
The only system that works more or less the way you are suggesting doesn't actually work that way either (yeah, I know that sounds weird, but I'll explain...) Audition has a center channel extractor - which does look at frequency content, but only in the center of a stereo signal. You can use this to boost or attenuate a vocal, but if you use it to extract the vocal completely, you'll find that inevitably, some of the backing will be there with it - because it's in the same place in the stereo field, and also because it has the same frequency content as the singing, so you're a bit stuffed. Even really clever and expensive software like Melodyne can't do this perfectly either, despite being better at the whole extraction process than anything else on the market.
So I'm afraid no - you won't be able to do it to a useable standard - at all.
1
Reply
AdChoices