A Vocal Chord (Part 1)
As you've probably already got your sampler playing drums, horns and strings, getting it to sing to you may sound deceptively easy. Tom McLaughlin explains there may be more to sampling the human voice than you'd heard.
If you're tired of hearing - and using - the same old vocal samples but have found making your own to be harder than it sounds, this two-part series on sampling vocals could add a new dimension to your sound library.
THE HUMAN VOICE is probably the most versatile musical instrument known to mankind. Not only is it capable of song and speech, it can also produce a multitude of sounds that are not part of any language as we know it. Clever beastie that he be, man can mimic much of the world around him, including members of the animal world, machines, weather, many musical instruments and sound effects.
Unlike the majority of sounds that make up our standard musical palette, the voice is extremely difficult to synthesise with any amount of realism using commercially available additive, subtractive, FM or digital synthesisers.
I thought I'd come up with some reasonable approximations using a Jupiter 8, Oberheim 4-voice, Moog series 12 and DX7 but these paled by comparison to the "oohs" and "aahs" of Tomita's recorded choirs. A more thorough understanding of the voice and its synthesis was in order... or so I thought. The bubble burst when I discovered Tomita "fine-tuned" his analogue vocal constructions with a vocoder.
Things changed quite dramatically when the sampled voices and choirs of the eight-bit Fairlight I, Emulator I and PPG Waveterm A appeared on the scene. Although they sound somewhat crude by today's 12- and 16-bit standards, and part of my brain still told me I was hearing electronically produced sounds, another part told me that these sounds were definitely of human origin.
Using sampling technology we can extend the range of the human voice beyond all known limitations (from sub-basement baritones to supersonic sopranos) or have our sampled vocalist sing ridiculously long notes without having to resort to mundane things like breathing. But before we get into the actual considerations involved in sampling vocals, let's take a look at the voice itself.
LIKE OTHER ACOUSTIC instruments, each voice has several distinct registers, each with its own characteristic tonal qualities. From low to high these are: the "growl register", "chest register", "head register" (or "falsetto"), and the "whistle register".
The growl register is difficult to sing a consistent pitch in, even for classically-trained singers. Using this register, sounds as low as only a few cycles per second can be made. It involves relaxing the vocal cords to such an extent that the only sound possible is a low growl of ambiguous tuning rather than a musical pitch. Funnily enough, it's easier to produce growl register tones while inhaling.
The chest register is the richest of all the registers and is what most people find themselves singing and speaking in. It uses a large percentage of the middle and upper body as a resonator to produce its distinctive "golden" tone.
The warmth and sonority of the chest register thins out and becomes more pure and "silvery" as singers reach their upper range, using less of the chest as a resonator and more of the throat and head cavities to produce falsetto or head register notes.
At the upper end of the falsetto register in women, children and some men, we find the whistle register, the shrill squeals sometimes emitted by young children while playing. Due to the skill involved, only the most talented of singers can produce these piercing, whistle-like tones with any degree of musicality.
Jimmy Somerville, Kate Bush and Olivia Newton-John have vocal ranges that border on the whistle register if my burnt-out tweeters are anything to go by, and bass singers in '50s vocal groups often reached down into their growl registers to hit notes so low that the pitch was inferred rather than actually sounded, but despite their obvious dramatic value, little use of the growl and whistle registers has been made in modern or classical music. Maybe with sampling technology enabling us to record difficult performances (filtering and manipulating them to taste) and replay them with the greatest of ease, we'll be hearing more of these vocal registers in film scores and popular music in years to come.
AS BEAUTIFUL AND musically useful as they are, the current palette of nicked and re-nicked "oohs" and "aahs" are beginning to wear a bit thin. The odds are that somewhere along the line these came from the Emulator and Fairlight stable of sample engineers, and my hat goes off to them for their contribution to sampling history, but there must be more to sampling than "lifting" ready-made multisamples. I find myself listening and playing "spot that sample" with TV commercials, and nine times out of ten I'll recognise the vocal pads as coming from the Fairlight, Emulator, Mirage or Prophet library.
Sure, it takes a bit of time and effort to record, sample, edit, tune, loop, map, trim and tweak a set of vocals, but at the end of the day it'll be yours - and there won't be another set exactly like it in the world. I like that idea.
Since ensembles seem always to be in demand, let's take a look at how you can create your own. But keep in mind that most of what follows pertains to solo vocal samples as well.
IT IS POSSIBLE to sample a voice by plugging a microphone into a sampler and going "aah", but I find it preferable to pre-record onto tape first. Among other things, I can "audition" recordings before I sample them, making sure enough level is getting into the sampler and, if need be, varispeeding the recording to alter the pitch. Besides, there's really no reason why a vocalist has to be present for what is basically pretty boring stuff for all but a sampling lunatic.
Those of you with access to a multitrack tape recorder with a varispeed control, a microphone and a mixer, have within your reach all the tools necessary for turning a solo vocalist into a vocal ensemble. The trick is to record multiple passes of your vocalist singing the same sound onto tape and record each pass at a slightly different speed.
Although "tracking" a vocal line several times at the same recording speed smooths out slight flaws and inconsistencies in the voice, on playback the voice comes out sounding like one big voice - nice as the effect might be, it still sounds like one singer.
If you vary the recording speed for each pass, you'll be shifting the formants of the singer's voice up or down, and the net result will resemble several people of different sizes singing together rather than just one voice. Samples of vocals recorded in this manner will also have a wider playback range before they start sounding strange, as the human ear and brain will have several sets of vocal formants to deal with.
To help further the illusion of an ensemble, try doing something different on each consecutive pass.
Things to think about are:
1. Using different EQ settings.
2. Changing mics and/or positions.
3. Having your vocalist sing with another coloration, in another style, or indeed like another person if possible and appropriate.
4. Varying parameters on the reverb if you're using one (it's a good idea to use one), or using a different "room" type for each pass.
5. Employing a digital delay, or dedicated effect unit, to produce a mild chorus or ADT effect. This will help to thicken and exaggerate your "ensemble". Care needs to be taken to keep any modulation to a minimum to aid future looping.
6. Anything else you can think of - every little bit helps.
WITH THE EXCEPTION of a few laboratory-quality microphones such as those manufactured by Bruel and Kajer, you'll rarely find a flat frequency response across the entire audio spectrum in a microphone - with good reason too. Just as it's impossible to find the "perfect" snare drum to do justice to every conceivable song, microphones can't be expected to record every instrument with equal precision.
In general you should use as good a microphone as you can get your hands on. Microphones intended for vocal use will have a gentle lift in the bottom end for warmth, and a boost in the upper end for presence and intelligibility, like the old standard Shure SM58. Most microphones have certain peaks and dips in their frequency response, some due to physical limitations, others actually built into their design, which, along with other factors, are responsible for their "character".
Give some thought to the overall sound you want your vocal samples to have. If you have a choice of microphones at your disposal, go through them all and get to know them if you don't already. Line a few of them up and route them through a monitoring system set dry and flat, and take a good listen to a mate or willing soul singing through each of them, making a mental note of each individual microphone's sound for future reference.
Have your singer move closer to and further away from each microphone to get an idea of how it deals with The Proximity Effect. As you get closer to most mics, the bottom end response is enhanced, and everything else for that matter, so it sounds fuller and richer. Although regarded as attractive on lead vocals, the fullness imparted by close-miking can often render group and background vocals very thick and mucky. This is especially true when multitracking solo or small groups of vocalists on tape to create the effect of a large ensemble as we're doing here. You might be surprised at how it all adds up.
As most vocals recorded for subsequent multisampling will be used in a supportive role in relation to the overall mix, I like to treat them as background vocals from the start in terms of microphone choice and placement. As I said, this involves them singing further away from the microphone than they would normally and equalising them to be somewhat lighter and thinner than a lead vocal. "Lighter and thinner" does not mean pushing more top end with your EQ but rather rolling off a little bottom end - try cutting frequencies below 200Hz.
If I'm after an airy, breathy vocal sound I'll go straight for a PZM microphone (the cheapo-cheapo mics sold by Tandy are good for this sort of thing) as their inherent frequency response and coloration usually leaves very little work to be done in the EQ department. For nice warm "humming" samples, I've had pleasant results from AKG D202s and Shure SM57s.
Although I haven't used every microphone around (who has?), here are some personal favourites for vocals, along with the qualities I associate with them:
- AKG D202: Deep and Warm
- Neumann U87, U47: Warm, round and Transparent
- Shure SM57, SM58: Rich and "Rocky"
- AKG 414, AKG 451: Crisp and clear
- PZM's: "Breathy" but thin-ish
CLASSICAL SINGERS SPEND much time and effort training their voices so that there is a smooth transition between the chest and head registers. The area where these two registers overlap is sometimes referred to as the throat register, but is actually a mixture of the above two registers.
For the most part though, you'll be working with vocalists who have a clear distinction between their chest and head registers. When making a multisample of a solo vocalist and, to a lesser extent, vocal ensembles, this transition area can stick out like a sore thumb.
Something that came back to me from my days in the school chorus was that if a vocalist sings an ascending scale, the chest register is carried up into what would ordinarily be the falsetto register. Conversely, if a descending scale is sung, the falsetto register is carried down into the range associated with the chest register.
Although this has a lot to do with the individual performer, it's a useful technique to be aware of when sampling voices (and certain other instruments, especially woodwinds, reeds and brass). When creating a multi-sample, a vocalist can be recorded singing both ascending and descending scales with the best takes of each note chosen for their contribution to a smoother transition between registers.
With group vocals, identical pitches can be mixed/merged from the ascending and descending scales, giving you not only a larger sounding group, but also a smoothing effect between vocal registers.
SINCE FEW SINGERS have perfect pitch, it will make their task infinitely easier if you provide them with a reference track to listen to on headphones as they do their overdubs, both for timing and pitch.
For pitch reference a good guide is a simple synthesiser waveform, maybe the first few harmonics of a triangle or heavily filtered square wave. But there's no reason why you couldn't use a vocal sample or sampled vocal waveform, as long as the pitch is consistent from beginning to end. If it takes your fancy, you could mix a little of this reference pitch in when taking samples of your recordings. It should help to beef up your ensemble, and will provide the ear with a central pitch if your vocalist's pitching is shaky.
The count-in to notes is a special case with vocalists. They need a pitch reference before they're to hit a note. To kill two birds with one stone, try using the pitch reference sound you're working with, and give it a percussive envelope with your VCA.
Make up a sequence of four, evenly-spaced pitched "clicks", a sustained note (at least two seconds long, preferably three or four) followed by a second or two of blank space for your vocalist to breathe (they need to do this from time to time). Copy and transpose this sequence, in semitone intervals, so that you end up with a master sequence, covering both ascending and descending scales, spanning your singer's entire vocal range. Commit this pitch and timing reference sequence to one track of your multitrack tape recorder and you're all set to start recording your vocalist. If you intend to do this on a regular basis you could copy this reference sequence onto cassette, so you don't have to set things up from scratch every time you want to record vocals for multisampling.
Although you won't end up using every single semitone from this sequence in your final keyboard mapping, there will be more than enough takes to choose from to whittle the lot down to those that sound best.
IT SHOULD GO without saying that one of the major prerequisites of a useful vocal sample is the absence of any vibrato. Vibrato-less samples generally give better "keyboard mileage", and besides, samplers more often than not have provision for adding vibrato later. This can be tailored to the project at hand rather than having to make do with what's there to begin with.
Unfortunately, losing vibrato is not the easiest of tasks for many vocalists. It's not as simple as leaving the mod wheel alone; serious vocalists spend a lot of time developing their vibrato and may rely on it to help shape their individual "sound".
The odds are that what you need as a sample will sound cold and lifeless to your singer and you'll have to spend some time coaching them to sing a clear consistent pitch while retaining the tone colour and characteristics of their voice. Some singers might not even be aware that they're employing vibrato at all. If you're having problems, you could take a sample of their voice and play it back to demonstrate how comic sampled vibrato can sound when played back as little as a third or fourth above the original pitch.
If you communicate your needs to your vocalist and make him or her feel involved from the start, you might be pleasantly surprised at how much more productive your session will be. Play back what you've recorded every now and than, and see if your vocalist is pleased with the results. If he or she feels they can do it better, give them another go - it's quite like conventional recording really.
Next month we'll go into trimming, looping and mapping these vocals. We'll also take a look at some of the more bizarre applications of the sampled voice. 'Til then, happy recording.
Feature by Tom McLaughlin
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!