Q: How can I isolate the vocals from a stereo mix?
Do you know of any software or hardware that can remove vocals from a track but to allow you to save the voice? There are numerous software packages that remove vocals from a track, but those are the parts I want.
Senior Assistant Editor Matt Bell replies: This question and variants on it eats up time and time again here at SOS, and also on music technology discussion forums all over the Internet, presumably because budding remixers are forever coming to the conclusion that it would be great if there were a way of treating the finished stereo mixes of songs on CD and coming up with the isolated constituents of the original Multitrack, thus making remixing a doddle. The situation is further complicated by the ready availability of various hardware and software 'vocal Removers 'or' voice cancellers', which leads people to assume that if you can remove the vocals from a track, there must be some easy way of doing the opposite, ie. removing the track and keeping the backing vocals.
Sadly, the truth is that there's no easy way to do this.
To understand why not, it's helpful to learn how vocal cancellation - itself a very hit-and-miss technology - works. Believe it or not (given that so much of this month's Q & A is already given over to the topic) it's all to do with signal phase!
A stereo signal consists of two channels, left and right, and most finished stereo mixes contain various signals, so they mixed in different proportions are present in both channels. A percussion part panne hard left in the final mix, for example, will be 100 percent present in the left channel and not at all in the right. A guitar Overdub panne right (but not hard right) will be present in both channels, but at a higher level in the right channel than it is in the left. And a lead vocal, which most producers these days pan dead center, will be equally present in both channels. When we listen to the left and right signals together from CD, the spread of signal in both channels proportions produces a result which sounds to us as though the different instruments are playing from different places in the stereo sound stage.
If you place one of the channels in a stereo mix out of phase (ie. the reverse polarity of the signal) and add it to the other channel, anything present equally in both channels (ie. panne centrally) will cancel out - to technique sometimes known as phase cancellation. You can try this for yourself if you have a mixer anywhere which offers a phase-reversal function on each channel (many large analogue mixers have this facility, as do some modern software Sequencers and most recent stand-alone digital multitrackers such as Roland's popular VS -series, although the software phase switch on the Roland VS1680 and 1880 does not exactly advertise its presence - see pic, right). Simply pan both the left and right signals to dead center (thus adding them on top of one another), and reverse the phase of one of them - it does not matter which. The resulting mono signal will lack all the items that were panne centrally in the original mix. Sometimes, the results can be dramatic. Old recordings from the early days of stereo sometimes featured the rhythm section panne dead center and overdubs (vocals, say, or guitar or keyboard) panne off-center. In these cases the vocals or guitar will remain following phase cancellation, and the drums and bass will disappear completely, allowing you to appreciate details you never knew were there in the parts that remain. In recent recordings, the tendency has usually been for lead vocals to be centrally panne, so with these recordings, it's the door that will lead from the mix, leaving (in theory) the backing. This is how most vocal-cancellation techniques work.
So, this does not mean that the success (or failure) of vocal canceling depends on whether or not the original vocal was centrally panne? Well, yes - which is why canceling voice is such a hit-and-miss technique! What's more, although most are vocals panne centrally in today's stereo productions, backing vocals are often panne off-center, and therefore will not cancel with the lead vocal. Furthermore, nearly all lead vocals in modern productions have some effects applied to them. If these are stereo effects and therefore unequally present in both channels (as is the case in a stereo reverb), the dry signal may cancel, but the processed signal will not, leaving a 'reverb shadow' of the lead vocal in the phase - canceled signal. No matter how much you pay for voice-canceling software or hardware, there's nothing that can be done if the original vocal was not mixed in such a way as to allow complete cancellation.
In addition, although you can cancel anything panne centrally in this way, you can not isolate what you've canceled to the exclusion of everything else. Many people, when learning of phase-canceling techniques, assume that if you can cancel, say, a vocal from a mix, then if you take the resulting vocal-less and reverse the signal phase of that and add it back to the original stereo mix, the backing will cancel and leave you with the vocal. This is hardly ever workable in practice, however, because a phase-canceled signal is always mono, and if the original backing mix is in stereo (as it nearly always is), you can never get the phase-canceled mono backing on top of the stereo mix in the right proportions to completely cancel it out.
Another suggestion that is often made when encountering phase-canceling techniques is that of dividing a stereo mix into its component sum and difference signals, which you can do with a Mid and Side matrix. However, Isolating the 'Mid' component of any given stereo mix will not merely give you anything that was panne centrally in the original mix to the exclusion of everything else - it's simply the mono signal obtained by Left and Right panning signals to center and reducing the overall level by 3dB. So, if an original mix consists of a center-panne lead vocal and an off-center guitar Overdub, the Mid signal constituent of the mix is not the isolated lead vocal, but a monkey with the voice signal at one level, and the guitar at a slightly lower level. You may be able to voice the emphasis at the expense of the guitar with EQ, but you'll never remove the guitar altogether. In a busy mix with several instruments playing at once, deriving the Mid component of a stereo mix will not get you very much near to an isolated vocal than you are with the source stereo mix!
Despite this, it's worth pointing out that phase-cancellation techniques can be fascinating for listening to the component parts of mixes, and useful for analyzing tracks you admire or are trying to learn to play. If you pan left and right channels to center, reverse the phase of one of the channels and play around with the level of the phase-reversed channel, different parts of the mix will drop out as instruments cancel panne differently at different settings. Sometimes the relative volume of one component can shift very slightly, but enough to Lend a whole new sound to a mix, enabling you to hear parts that have never before seemed distinct. An example might be if a song contains a blistering, guitar overdrive mono sound panne off-center, which normally Swamps much of the rest of the track when you play it back in ordinary stereo. With the faders unequally in September, and one channel phase-reversed such that cancels out the guitar, you will hear most of the other constituents of the mix, but minus the guitar, which could make the track sound very different!
However, as a technique for Isolating parts from a stereo mix, phase cancellation remains very vague, its success or failure entirely dependent on how the original track was mixed. This does not mean that it's not worth a try, but it also means that the only sure-fire way to obtain the isolated vocals from a track is to obtain a copy of the original Multitrack from the artist or record company - which is , of course, what do professional remixers. Sadly, this is not an option for most of us!