How to extract vocals from an audio recording?


savemesample
The sound sample above is a example of KnockOuted vocals of OMD's "Save Me" featuring
Aretha Franklin
. First you hear the original, then the extracted vocals.

I'm very much into making fan remixes and mashups. Sometimes people ask me how I extract the vocals from the backing instrumentals in a recording. Well, there is one short and one long answer to that. The short answer: visit
Acapellas4U http://www.acapellas4u.co.uk/download_search.php . At this page there are loads of acapellas of various artists from the 1950's to present time! A lot of them are studio acapellas i.e they are taken direct from the master tapes. How the site got their hands on that is more than I can comprehend, but it is a useful tool for all dj's and basement musicans all over the world.
However, some acapellas are so called
DIY acapellas - Do It Yourself acapellas. They are not as perfect as the studio acapellas, but usually good enough to work with. That leads us to the long answer of the question: how is it possible to extract voclas from a mix?
Lets be a bit technical.
I extract vocals with
Audacity http://audacity.sourceforge.net/download/ with the plugin KnockOut added http://www.freewebs.com/st3pan0va/ . Both are freewares.
To use KnockOut I need to line up a exact instrumental version of the song to the original. If I do not have that it can work any way if the song is based on a sampled groove. In such case, I use to make a loop of a instrumental part of the song and line it up with the original.
Sometimes the back track of verse 1 is equal to verse 2. In that case I line up verse 1 with verse 2. And it have to be exactly matched!


Step-by-Step how to extract vocals using the KnockOut plugin. Click the picture to enlarge.

Next, I split the stereo tracks on both the original and the instrumental/groove/verse2. In Audacity, I click on each track header and choose "Split stereo track".
Now you have four mono tracks. Organize them as below:

Left Original
Left instrumental/groove/verse 2

Right Original
Right instrumental/groove/verse 2

I make the both left mono tracks to a stereo track by clicking on the Left Original header and choose "Make stereo track". I do the same for the Right Original.
Now it's time to use the KnockOut plugin. I mark all channels and click the effect drop down menu and select KnockOut. Voila - acapellas is made for the left and right channels! You will probably hear a muted version of the instrumental in the background, but buried in a new mix you hardly hear any of it. There are several reasons why the extraction of the acapella are successful or not:
- Sound matching vocals - instrumental. The more exact - the better acapella.
- The sound quality of the original recording. The better bitrate - the better acapella.
- The original recording equipment. Digital recordings are better than old analogue recordings.
- The mastering compression. Especially radio edits of singles are usually harder compressed in the mastering, and that makes it harder to get a successful acapella



greensample
Listen to the sample of Green by OMD (first with vocals extracted
with KnockOut, then the whole mix). It is a example of a good
audio recording:
good sound quality, digitally recorded and not to much compression.


To clean it up, I (again) split the stereo tracks, remove the right channels ( they have become silent in the process) and make the two acapella tracks a stereo track.
Still, there are some question marks left. What if the song isn't based on samples and made before the multi track recording system was even invented? I do not have the answer to that question, but someone certainly has! Listen to
Saveyourmix' vocal extraction of a Elvis' "That's alright Mama" on Youtube:
http://www.youtube.com/watch?v=QgiqQD9Tltg
Incredible!
I think this Saveyourmix guy has used a autotuning software called
Melodyne ( http://www.celemony.com/cms/index.php?id=demo_editor ) It is not free, but there is a trial version available. With this software, it is possible to edit individual notes in polyphonic audio recordings. By simply deleting or muting the undesirable notes it is possible to just leave the vocals. But it’s not an easy task. I have tried it myself, but it always end up sounding crap. It takes a talent like Saveyourmix to achieve that!