Acapella Creation Tutorial

[Tutorials]

Step 1 : Instrumental Version

Find an instrumental version of the track. CD Singles are the best source, as the digital recording & pressing techniques are more likely to give a perfect copy of the backing track, which is something that you need.
Mp3s are not likely to give good results as the compression method removes some of the audio data that may be required. Recording from vinyl may work if you do both recordings under exactly the same conditions (any pitch variation will reduce the effectiveness of this technique). Just rip the audio tracks to wav files, or another uncompressed format, .AIF etc, using an Audio Extracter/Ripper. The basic idea relies on the linear superposition of audio. Which is simply: that if you have take a sound, any sound at all, and add it to the inverted version of itself the result will be silence.

Note: Occasionally the instrumental on the CD Single will have a slightly different arrangement or structure, unfortunately in this case you're stuck. A rough guide when you're considering buying that CD single is "Are the two tracks exactly the same length?" if they are, then there is a good chance the engineer just muted the vocal track when they did the instrumental, which could result in a nice clean acapella.


Step 2: Matching Them Up

Now that you have the Instrumental Version its a matter of matching the two songs exactly. i.e. to the sample. The best thing for this is a wave editor like Soundforge, Peak, Cooledit etc.

Radio version vs Instrumental

When loaded into your wave editor the two tracks should look pretty close except for the vocal sections. Thats a good start. By zooming in on the first beat of the Radio Version you get a better picture of the shape of the waveform. You want to find a distinctive little waveshape that you can then look for in the instrumental (or vice versa).

After you find the same section in the second track you can start trying to line them up so that the spikes fall on EXACTLY the same sample position.


Step 3 : Mix 'Em Together

The rest is rather simple, just invert one of the tracks, say the instrumental, ("Select All" then apply an "Invert Waveform" function or similar) and mix the inverted instrumental with the original aligned radio version the result may be a pretty clean acapella.

To Mix the two:

We will use Soundforge for this example.Select the entire song (try and find a "Select All" command or use "Ctrl+A" or "Option + A"). Copy that into the clipboard. ("Ctrl + C" or "Option + C"). Select the other version from the begining of the track and paste mix ("Ctrl + V", "Option + V"). You may need to adjust the repective volumes of the two as there may be differences in the mastered loudness.
The idea is that now that they are aligned it is very easy to do a paste mix where the two waveforms will fall directly on top of each other (but inverted) and hence cancel out. This will leave only the difference between the two (the vocal)

Most likely still be some portion of drums and other instrumentation in the background throughout the acapella. This is most likely due to post production mastering differences (multiband compression, maximising reverb etc.) between the instrumental version and the radio edit. We've also noticed there is often some nasty digital noise, its low in the mix but can be heard in the straight acapella. We believe it is due to dither (which is the process of purposefully adding noise to digital recordings to smooth out the track) as it should be random each time so won't cancel out.


Rumoured Methods

1. EQ out the Backing: We've often heard people talk about EQing the backing track out of the mix, put simply it doesn't work. It can't be done without destroying the vocal as well. The tonal range of the human voice is also within the range of most musical instruments & therefore it will never work.

2. Karaoke Versions: There was an old school method to create a vocal free version of a song. Which relied on the notion that the left and right channels of most instruments are well out of phase however the vocal as it is mono and mixed center is not. Thus mixing say, the right channel with an inverted left will result in a vocal free version. This will give you an "instrumental" or a rough one. You can not use that in this process: some simple algebra will explain why it will not work.


A little EQ afterwards will help clean it up and maybe a touch of multiband compression to emphasis the vocal and add some dynamic noise reduction.