Removing voice

I’m trying to remove the vocals from mp3 so as to create an instrumental to use at a special occasion.

Not wanting to sound negative, but in most cases vocal removal either does not work, or sounds dreadful. Having said that, there are some suggestions about how to attempt vocal removal here:
If vocal removal fails (as it probably will), have a look on Google/Amazon/etc to see if a karaoke version of the song is available for purchase.

There is a YouTube instructional video about how to do this job in Audacity – the hard way. The one I saw takes you carefully through all the individual steps of splitting the tracks, phase reversal, etc. etc. etc.

Effect > Vocal Removal has packaged all those steps into one tool and one pass for you.

Most songs do not lend themselves to vocal removal. This is one of mine that didn’t work very well. The clip plays After and then a little Before for comparison.


Doesn’t Sound too bad Koz. The piano Comes fine up in gain.
When you sing along such a Version, your voice is sometimes able to mask all the remaining lead voice artefacts.
And when you sing as badly as I do, it will be advantageous to have sum original remaining…

Please note the bass and drums vanished, too. This clip illustrates almost everything that can go wrong except I started with a Very High Quality Stereo original song. No compression artifacts to struggle with and no accidental mono download.

People who post these are almost always after the theatrically perfect karaoke, not a missing instrument version with the original singer in deep echo. That song is only likely to happen by writing a check to a music supplier.

We also point to the song that the teacher used. If you use that exact song, your karaoke will come out right as well. What you can’t do is by extension apply this technique perfectly to all songs.


You can somewhat preserve the bass with the frequency band choice.
I am not sure if the current implementation is ideal though.
It uses a filter with a roll-off of 48 dB/octave.
One Needs enough head room when applying this filter (the Output is sometimes louder than the Input).