Faking Ensemble Backing Vocals

Tips & Tricks

Technique : Effects / Processing

Paul White comes up with a few tips for using the processing power of your computer audio sequencer to turn one or two voices into a complete backing ensemble.
 
If you want rich, lush backing vocals, few things beat having a small or medium-sized group of people with different voices singing in harmony. It can, however, be hard enough to find and record one good vocalist, let alone several: all too often, practical considerations mean that backing vocals have to be recorded by only one or two singers. Ever since multitrack recording was invented, singers have been double-tracking or layering their voices to create a fatter sound, but these techniques still don't create convincing ensemble effects because no matter how many overdubs you use, the character of the voice will always be the same. In a real vocal ensemble, people with different vocal characteristics, slightly different pitching and slightly different timing combine to create a rich, natural vocal texture.

To simulate this using technology, it's necessary to emulate all these variables if the end result is to stand any chance of being convincing. If you can get two different people to sing the same part, then the following techniques are likely to be even more successful, but even if you're doing this with just one voice, the results that can be achieved are often quite impressive.

Clone-A-Vocalist

The first decision to make is whether to treat a single vocal line or whether to sing the same part on to several tracks and then process these separate tracks to create the various layers of the ensemble. The advantage of singing the same part on to several tracks is that no matter how good a singer you are, there will always be slight timing and pitching differences that you can exploit to create your fake ensemble. On the other hand, if you find it difficult singing the same part twice, it may be best to start from a single vocal line.

The technique used here is to make copies of the original track on to new tracks, then use the processing functions available within your computer sequencer to treat each of the copied parts so that it differs from the original in pitch, timing and timbre. This means each layer must be processed using slightly different parameter settings, and the more layers you use, the richer the ensemble sound will be. If you're limited by the number of tracks available, record and process a few layers at a time, then bounce them down to one track (keeping the original part or a copy separate) before repeating the process for further layers.

Processing Types & Effects

What processes you use will depend on what outboard processors or plug-ins you have available, but fortunately, there are several viable alternatives. To create the illusion of a different voice, check out the features available in the pitch-shifting section of your MIDI + Audio sequencer. Most modern systems allow formant-shifting as well as pitch-shifting, and a slight change of formant frequency can radically alter the timbre of the voice. At extreme settings, you can even make male vocalists sound female and vice versa. By shifting the formant structure of your voice up or down from the original, the copied tracks you create should give the the illusion that a completely different person is singing each part. Logic Audio users can use the Pitch/Time Machine section of the Digital Factory to do this. The trick is not to add so much formant shift that the voice sounds unnatural.

Depending on the system you have, you may also be able to add some pitch-shift at the same time, and though a fixed shift won't sound quite like the natural pitch variations that occur when two people sing in unison, it's still quite usable. Try using shifts of between +/- 3 and 12 cents, and make sure you have roughly the same number of parts singing flat as you do singing sharp so the overall pitch averages out as being in tune. If there's an audible chorus effect, you've probably gone too far.

A better way to simulate natural tuning differences is to use a software plug-in that can correct pitch, such as Antares Autotune VST (or Autotune TDM for Digidesign users). Playing the original take alongside the corrected version will result in small, random pitch differences, depending on how precisely the original part was sung. By creating different tracks, each using a different response speed and correction percentage, you'll end up with several tracks that have apparently natural variations in pitch. External processors such as the Digitech Vocalist or the Lexicon PCM80/81 (Pitch card needed for PCM80) can also be used to pull the pitch of a vocal performance closer to a chosen musical scale. Alternatively, you could use a sampler to replay each part, using a little pitch-wheel modulation to introduce random pitching differences.

The small timing differences between singers can be emulated to some extent by adding small amounts of positive or negative delay to the various tracks before mixing them - a spread of around 50 to 150mS should be adequate depending on the vocal style. If the attack of the first syllable of a phrase is too pronounced, the sum of the various delayed versions might sound too ragged, in which case it should be possible to use your sequencer's graphical level envelope facilities to add a slower attack to some of the parts. The same goes for endings - fade out the final syllable on a few of the more widely spaced tracks and the ending will tidy up nicely.

The Final Touch

So far we have created a number of copies of the original vocal track or tracks, but used formant-shifting, pitch-shifting and delay to make them sound like several different people singing the same part. The effects can be further enhanced by using an appropriate reverb setting; if you don't want an overtly wet sound, a short ambience setting with plenty of pronounced early reflections can help add to the overall density of the sound. Longer reverbs can benefit from a little chorus or swirl, either within the reverb algorithm, if such a facility is provided, or by patching either a chorus or a pitch detuner before the reverb input.

The level of success that can be achieved depends on the processing power you have at your disposal, and on your patience. It's also generally true that having two or three different voices to start with makes it easier to build up complex choral sounds than working entirely from a single vocal line, but not every song demands backing vocals of a choir-like density. The key, as always, is to allow yourself time to experiment, and make a note of what works.

Published in SOS May 1999