Amazing Multimedia   
info@amazingmultimedia.net   
808.676.0600   

Let's start off with the bottom line: in most cases you can't do it.

Here's an analogy - you use eggs to bake a cake.  Once you mix those eggs with flour, butter and whatever else and bake that cake you can't take the eggs back out.

You'll find lots of ads on the internet for software, plug-ins and hardware that say they can remove vocals.  Then, if they're honest, they'll be some small print that says, "Your mileage will vary".  The fact is, no matter what you use, they all work the same way and all get the same lousy results.

The sounds you listen to, whether they come from a record, tape, radio, CD or someone talking, are all a mixture of frequencies.  People's voices are in a certain frequency range, the car horn tooting down the street has its own frequency, the wind whistling through the tree, the laughter of children and the sounds of a musical instrument.  The only way to remove vocals is to subtract the vocal frequencies from the material.  Common sense tells us that's "iffy" at best.  Musical instruments have an extensive range of frequencies as it slides up and down the scale of notes it can play. Some of those frequencies are going to be the same as the frequencies of people's voices or singing.  That's the way it is.  Can't be changed.  Trying to remove a vocal by using a plug-in to cancel that frequency will affect all the musical instruments that share that frequency.  Most commonly, vocals (or any sounds) are removed by taking the right channel signal and subtracting it from the left channel signal (or left from right).  In other words you are canceling out frequencies that are the same.  And that's why trying to remove vocals really doesn't work.  Before you shoot us an e-mail, yes, we can do the same thing by taking one channel of the audio and inverting it 180 degrees out of phase to the other channel. If the vocal is the same level for both channels, it's gone.  Unfortunately so is everything else if we have a mono record.  Okay, no vocal removal from mono mixes.

If you have a stereo mix you might have some success - depending on a few things:

1) the vocal has to be dead center
2) everything else must not be dead center

Dead Center: When  the level in the right channel is exactly the same as the level in the left channel.

The bass guitar and kick drum are usually dead center because these are low frequency instruments and low frequency instruments are referred to as "non-directional".  Non-directional means your ears can't tell if the sound is coming from the left or right so they are mixed dead center because it's easier.  Other instruments like guitars, keyboards, drums are not panned dead center. This means that they are louder in level in either the left or the right channel or one of the back channels if you're using some version of surround-a-sound.  So, in a perfect world we would subtract what is "common" (or the same level) in the left channel from the right channel.  If the vocal signal is the same in the left and right channel and we subtract them, then this signal (the vocal) will disappear from the mix.  The bass guitar and kick drum, which were also dead center, usually have enough harmonics spilling over to the other channel we can restore them.   Alas, the world is not perfect.

Why you won't get the vocal removed

Most vocals are blended with reverb and/or other effects. In most cases this is designed to fatten and improve the vocal sound.  (Think Ashlee Simpson)  Almost all vocals are recorded as a single tract, or in mono.  But when the engineer adds reverb (or another effect) it becomes a stereo track. That's because reverb is a lot of echoes happening at different times, making the echo sound of the reverb different in the two channels.  So, the reverb part of the vocal is not the same in the left and right channels and therefore will not disappear when you do the left channel minus right channel.  Given that, the dry vocal (or non-reverb portion) may disappear, but you're left with something sounding like someone singing in the shower or at the bottom of a well.

There's more.  Over the past few decades many sound engineers would record 2 vocal tracks and blend them together in the mix to give the vocal a fuller, more rich sound.  Phil Spector made this famous in the 60's with his "wall of sound".  Commonly, one vocal track is panned slightly left of center and the other slightly right.  You will hear these as identical and the vocal seemingly dead center.  They are not.  The signal levels of the vocal are not the same in the left and right channels and you will not be able to remove it completely.  Depending on the signal difference, what remains will vary between barely there and no change at all.

The same thing happens if there's only 1 vocal track and engineer panned it slightly off of dead center. The vocal will be reduced but not eliminated.

Back to our Perfect World

If your karma is great and your yin & yang in balance you might get lucky.

This would be a mix where the vocal was panned dead center with no effects blended with it.  All of the other instruments would be panned far right or far left.  If all that happened then you'd have a pretty good chance of removing the vocal and being left with a good instrumental.  Your only problem is that engineers haven't mixed this way in 40 years.