D.I.Y. a cappella

From Wikipedia, the free encyclopedia

DIY a cappella (DIY representing Do It Yourself) refers to the procedure of manually separating a vocal-only portion of a music track from its instrumental counterpart; also known as ‘vocal extraction’. Conversely, a DIY Instrumental is achieved by performing the same procedure inversely.

There are three primary methods used to achieve a vocal-only music track from a fully-mixed (a cappella & instrumental) version.

1. Phase Cancellation Method (superimposing two waves)

An instrumental version of a song’s phase is inversely mixed with an original version – resulting in a pristine cancellation of instruments.

Method:

"Get an instrumental of a song, invert the phase (sometimes referred to as flipping), and mix it with the original. If done right everything besides the vocal is cancelled out. You can do this in any sequencer like cubase or acid. The instrumental must be exactly the same time/pitch however, and mp3s might not work if they are badly encoded. Zoom right in to see the 2 waveforms next to each other (look for kick drum hits) and line them up."

2. 'kn0ckout' Method

A VST plugin (compatible with Adobe Audition) called ‘kn0ckout’ is used to spectrally subtract the instrumental part from the instrumental+vocal part.

Method:

Using 'knock0ut' (http://www.freewebs.com/st3pan0va/) as a plugin in a VST compatible audio editing platform, you can spectrally subtract one piece of mono digital audio from another. If you have a with-vocal and instrumental-only version of a piece of music, you can use kn0ck0ut to subtract the instrumental track from the instrumental+vocal track, and possibly substantially isolate the vocal track.

Kn0ck0ut processes its input in the audio frequency domain, and so is different to the phase-cancellation method, which works in the time domain. What this means in practice is that you can still achieve some sort of result even if the instrumental part you have is similar to, but not sample-for-sample identical to, the instrumental part underneath the vocals. For example, in dance music - which often uses repeated instrumental sections - it is often possible to find an instrumental section in a with-vocal track that can be copied and looped, and then subtracted from a with-vocal section using kn0ck0ut. The quality of the result depends upon how closely the instrumental loop matches the intrumentation under the vocal section, and on how well the two parts are synced up, as well as on the limitations of the processing method itself (see below).

Kn0ck0ut can also extract 'centre-panned' material from a stereo track. On many early (<1970) stereo releases, and much more rarely on recordings made since then, the vocal is the only part placed in the centre of the stereo field. If (and only if) this is the case, kn0ck0ut's extract centre feature will return only the vocal as a mono audio stream.

The quality of the result is however limited by both the nature of the input material and by the nature of the spectral processing (FFT followed by iFFT) employed which introduces sonic artifacts into the output.

With amenable source material, it is possible to create a usable 'DIY acapella' by this method. It is, however, far from being a complete or reliable solution to the problem of extracting vocals, or other instruments, from digital audio.

3. Vocal/Instrumental Filter

These methods vary in the procedure(s) applied to the audio wave. Programs such as ‘Goldwave’, ‘Cooledit’, 'Audition', and ‘Soundforge’ feature ‘vocal removal’ and ‘instrumental removal’ filters. These filters rarely produce adequate quality Acapellas and Instrumentals. Their physics usually involve inverting one audio wave’s channel and mixing it into the un-inverted channel.