Vocal Production

Just how do you achieve that extra bit of professional polish for your vocal parts? Find out with our in-depth guide...

Any attempt to explain the process of vocal production in its entirety, with all the different genres, styles and subjective opinion that this would entail, is, to some extent, doomed to failure — there's just no way to pack all that we have to say about this topic into a magazine article. But so critical is the role of the vocals in most contemporary music styles that we thought we should give it a go anyway — so whether you need help with arranging or performing, recording or mixing, there should be something here for everyone.

One notable contemporary production fashion is to keep vocals sounding relatively forward and free from obvious reverb without allowing them to dry out altogether, and that's one area we'll focus on. However, along the way we'll be throwing in a few other tips and tricks that will allow you to add variety to the vocals in different parts of a song while still retaining the illusion of minimal processing. I can guarantee that you won't learn everything there is to know about vocal processing, but at the same time I'm sure that you will take away something useful that can be applied in your own mixes.

The Low-down On Popping

Pretty much everyone knows that using a vocal mic up close without a separate pop screen of some kind is likely to result in audible popping on those plosive 'B', 'P' and 'M' sounds. What may be less obvious is that a significant amount of low-frequency energy from these sounds still makes its way through a typical pop screen, and though it's usually too low in level to be audible, opening a spectrum analyser plug-in on your vocal track can show up considerable activity down as far as 20Hz, or even beyond. Although you may not be able to hear these low frequencies, they eat up valuable headroom and occasionally conflict with the legitimate low frequencies in your mix from other instruments.

A low-cut filter, either on the mic or on the preamp (most are 80Hz, 18dB/octave) employed at the recording stage will help, but even that may not entirely cure the problem — so additional low-cut filtering of the vocal track using a similar filter while editing or mixing may be needed to adequately suppress those unwanted lows. In most cases, you can afford to increase the low-cut filter frequency until you just start to hear it affecting the tonality of the voice, then back it off again slightly so that the filter is set as high as possible without compromising the low end of the vocal sound.

Low frequencies can still make it through a pop screen and a low-cut filter — either on the mic/preamp or during the mix, or both — will help clean things up before you start performing other processes. In this screen shot, a low-cut filter is just one part of the EQ in an instance of the SSL Vocalstrip plug-in.

Squeeze Me

Pop production makes regular use of compression on vocals, not just to even up the level but also to increase the average vocal energy, to help vocals stay on top of a busy backing mix. However, over-compressing a vocal can stifle the life out of it, so it's often better to use level automation to fix any obvious level disparities and, where necessary, to ride the overall vocal level to suit the dynamics of the song. This leaves the compressor with less work to do, so you can adjust it to give the most artistically pleasing sound rather than having to rely on it to level the vocal unaided. A moderate ratio of between 2:1 and 4:1 with an attack in the region of 10ms and a release time of 50-100 ms usually works fine as long as you don't apply more than 5dB or so of gain reduction on the peaks. Rock and urban music styles can usually stand rather more compression, but for a natural sound, keep the amount of moderate. After all, the overall mix is likely to be subjected to even more compression at the mastering stage, or even before then, if you're mixing through a bus compressor.

Bear in mind that your DAW's channel fader is usually set up so that it governs the level after any insert processors like compressors or EQs in the signal chain. If you want the level automation to affect the signal going into the compressor, you'll need to edit the levels on the waveform, automate a gain plug-in before the compressor, or automate the compressor's input level. Alternatively, you could place the compressor in a post-fader insert slot if your DAW allows that, or route the corrected channel to a bus, where you then add the compression. There's nothing wrong with automating the level after the compressor, but do be aware that the two approaches will give different results, as the compressor is working on different signals in each scenario. In practice, very often you'll end up using both tactics.

I often use parallel compression (the signal is split in two, and one treated with a compressor while the other is left untreated) on lead vocals, but the compressed signal has to be mixed in sparingly to avoid the result sounding overblown and congested. By feeding both the parallel compressor and any necessary reverb or delay effects from the vocal channel's aux sends, it's possible to keep the parallel compression contribution completely dry. When it's mixed in with the channel signal and any added reverb and delay effects, this helps add weight without losing the focus of the sound. The usual rules of engagement apply when setting up the parallel compressor: use a high ratio and low threshold to achieve gain reduction values in excess of 20dB and set fairly brisk attack and release times so that gain pumping is actively encouraged. You only need add in around -20 to -15dB of parallel compression to notice the vocal gaining weight and authority.

EQ Or Multiband Compression?

Other than the low-cut filtering that I mentioned earlier, there are no golden rules as to how much EQ a vocal will need to make it sit comfortably in the mix. Much obviously depends on the character of the voice, and that of the microphone and (rather less) the mic preamp that were chosen to record it.

This screen, from a Tinchy Stryder session, shows typical EQ for a vocal part. Don't try to copy the settings exactly, as they'll be different for every part and track, but do note the small notches to remove resonances, the attenuation of the lowest frequencies, and the broad, gentle nature of the boosts.

Personally, I prefer to add as little drastic EQ as possible. Occasionally I'll use a gentle notch to tame a low-mid 'honk', but where the singer's voice starts to sound aggressive on higher notes or on louder sections, I find that using just one band of a multi-band compressor is often enough to tame it. You can use a conventional EQ to sweep through the vocal spectrum to see where the harshness is occurring — which will usually be somewhere in the 2-5 kHz range — and then set your compressor band to enclose the same frequency range, bypassing all the other bands. With the track playing, adjust the compressor threshold so that significant gain reduction occurs only during the louder, more strident sections. Again, a 4:1 ratio or thereabouts with a 10-20 ms attack and a 50-100 ms release time should get you close to where you want to be, and in most cases 4-6 dB of gain reduction will be enough to polish the vocal sound, without making it sound processed. You can use exactly the same technique for de-essing if you don't have a good de-esser plug-in — just use your conventional EQ to locate the offending frequency range, then set a compressor band to deal with it. If the vocal needs brightening, a touch of 8kHz high-shelving boost often does the trick. For a more detailed look at de-essing, check out our in-depth article in SOS May 2009 (/sos/may09/articles/deessing.htm).

Space

I have a theory that back in the late '70s and '80s, when good digital reverb units were still incredibly expensive, engineers and producers often piled on huge amounts just to get their money's worth out of their investment! Today, reverb may still occasionally be used 'to excess', but there's a more general trend in current pop and rock music production to aim for a less obviously wet vocal sound, while still creating enough of a sense of space to help the vocal sit comfortably with the backing. Certainly, high levels of reverb have the psychoacoustic effect of pushing vocals further back in the mix unless the arrangement leaves plenty of space for the vocals, but there are simple strategies that can be used to keep the vocal sounding more intimate.

To recreate the sound of a small club, there's nothing, really, to beat a convolution reverb loaded with an Impulse Response taken from the desired environment, but one of the reasons why the plate-reverb sound remains so popular for more general treatments is that it adds a desirable degree of wetness and gloss without triggering your brain's 'room-recognition circuits' — so you're not thinking cathedral, car park or concert hall, you just hear a musically sympathetic reverb. Real plates are scarce these days, but pretty much every digital reverb has a plate emulation, and Universal Audio's EMT Plate 140 plug-in for their UAD platform comes incredibly close to the sound of the vintage hardware.

One of the oldest and best-known methods of keeping the vocal sounding forward is to separate the dry sound and the following reverb using a delay — something we call pre-delay, which most reverb plug-ins have built in. Pre-delay times can be anything from a barely-noticeable 30-40 ms to well over 100ms. Pre-delay has a real-life counterpart, because if you're standing close to someone in a very large space, you hear their voice coming over dry and bright, then when the sound of their voice bounces back from the distant walls, you hear the reverb a fraction of a second later.

The screen shot (above) shows the simple quarter-note delay (via Pro Tools' Mod Delay II plug-in) used by Young Guru (right) to treat Rihanna's vocals on Jay-Z's The Blueprint 3 album.

If the reverb is set to be tonally bright, it sounds rather obvious, though some styles make use of this fact (think George Michael!). A less obtrusive approach is to roll off some high end from the reverb, so that it tops out at perhaps only 3-4 kHz as, in combination with pre-delay, that will keep the vocal sounding fairly up-front and intimate. It's almost as though your brain processes the presence of reverb without you hearing it in a conscious way, as you would with a bright reverb sound.

I also tend to roll out the lows below 150Hz or so, to stop the reverb tail sounding too cloudy in the lower mid-range. An old-school algorithmic reverb works well for this, and sometimes turning down the density and diffusion to produce a coarser-sounding reverb ends up being more flattering than a modern, dense reverb or convolution room sound.

You'll hear this kind of smoothed-over reverb used by bands such as Kings of Leon, The Killers and Red Hot Chili Peppers, to keep the vocal sounding forward and fairly dry but without allowing it to sound too stark, though Kings of Leon sneak in a little delay too, on some tracks. Many Coldplay tracks also tend towards a dry vocal sound, but just occasionally they'll fall back on an old-school, slightly more obvious reverb treatment. While many of the records made towards the end of last century very often relied on a generous application of reverb throughout, today's producers are much more likely to vary the vocal treatments from track to track. Listen to a recent Strokes album, for example, and you'll hear a range of treatments, from a dry, short reverb with pre-delay to John Lennon-style slapback echo or pitch-detune thickening.

There are some very successful current acts that rely on a more generous application of vocal reverb, Fleet Foxes being a case in point. Most of their tracks seem to use an algorithmic reverb with pre-delay and a fairly long decay time, quite possibly a Lexicon, with some top rolled off, but not so much as to take all the sparkle out of the reverb. Indeed, they seem to have taken almost an old-school approach to reverb, but if it works, don't knock it. Bands as diverse as JLS and Arctic Monkeys take a similar approach, often combining reverb with delay to create a big sound that wouldn't be out of place on a vintage Duran Duran track. Both Adele and the late Amy Winehouse, on the other hand, often seem to use a different reverb for almost every track, ranging from a fairly lively and quite obvious plate-reverb-plus-delay treatment to a very intimate and barely noticeable room ambience.

A useful tactic for creating a forward, but not too wet, vocal ambience is to use a reverb that has a balance control for the early reflections and the late reverb tail. If you sway the balance in favour of the early reflections, you can create a sense of ambient space, but without adding any appreciable reverb tail. Again, turning down the density and diffusion controls, and also adding pre-delay, can help give vocals a sense of place and closeness without washing them out with reverb. If the reverb has a size control, this will adjust the spacing of the early reflections, giving you more control over the character of the sound. However, you do have to avoid adding so much that you give the game away — it's surprising how little you need to add to kill the dryness and make the vocal sit in the mix. Some reverb plug-ins come with dedicated ambience programs, and these are worth trying — and even a gated reverb can work on vocals, as long as you keep it low in the mix.

Sometimes you can forget the reverb altogether and use delays instead. I've had success using a stereo delay, with one side set to around 100ms, without feedback (to create a pre-delay/doubling kind of effect), and the other set to several hundred milliseconds, often at or close to the song's tempo, with enough feedback to produce three or four audible repeats. Set the delay balance with the rest of the track playing, as what sounds like a lot of effect when soloed can almost disappear when the track is playing. A modest amount of stereo delay can also work well in combination with a mild sprinkling of reverb.

Tuning & Other Tricks

We've covered the use and abuse of automatic tuning plug-ins many times, including using a relatively slow retuning speed on a slightly delayed copy of a backing vocal to create a pseudo double-tracking effect when both are played back together. You can also create surprisingly subtle main vocal effects, though. Try feeding the delay or reverb send via a pitch-correction plug-in, set to the scale being sung, and with the correction speed cranked up full. This way, the original retains its pitching while the delay and/or reverb effects are fed from a hard-tuned signal. Balance this effected signal with the main dry vocal and you have a new twist on an old effect that sounds far more subtle than you might imagine.

These more creative treatments often crop up in dance music production. Other favourites in dance genres include feeding a delay via a bit crusher, granuliser or stutter plug-in, so that the delays sound dramatically effected while the dry part of the vocal remains intelligible. There also seems to be more chance to use excess levels of delay and reverb as spot effects too in these genres — though there's obviously plenty of scope for dramatic spot effects in modern pop mixes, too.

Ringing The Changes

As I've hinted already, setting the right reverb or delay balance can be quite difficult, because in a busy mix those reverbs and delays may need to be pretty loud to be audible at all. However, if you hit a drop-down section, such effects will become exposed and can suddenly sound overblown. The simplest remedy is to automate the effect send levels, so that you have less audible effect during exposed sections of the song. In practice, though, I'll often move the quiet section vocals to a new track and then apply a different treatment to them. For example, the early-reflections ambience treatment can work well on exposed vocals as long as you don't use too much of it. You can then apply a stronger reverb or delay for the louder song sections.

Yet another way to beef up louder song sections or choruses without adding more delay or reverb is to copy sections of the vocal line to a new track, then add a little subtle pitch-shift (5-7 cents is usually enough) to the copied section, combined with a few milliseconds of delay — again, a method of faking double-tracking. You'll also find that when you start stacking up vocal parts in this way, you need far less reverb or delay to make them sit comfortably in the mix, as the more layers you have, the richer and more textural the sound becomes. Using subtle pitch-shift effects to thicken a vocal may seem pretty old-hat, but the real trick is picking just the phrases you need to emphasise and applying it only to those, rather than trying to double up the entire vocal.

Producer Tony Maserati: "In the analogue days I would come up with one [vocal] effect for the entire song. Now, I'm putting different effects on individual lines and even individual words.”

The Effect Balance

Combining any of the effect treatments outlined above with parallel compression (discussed earlier in the article) offers the advantage that you can keep the parallel compressor signal completely dry if you want to, or treat it with a completely different reverb or delay. When you blend the dry compressed signal with the main channel signal (and its associated reverb or delay effects), this has the effect of adding density but also of drying up the combined sound slightly more, in a way that still allows it to retain a sense of space. That's because the signal peaks are still carried mainly by the original vocal track, with the parallel compression just pushing up from below.

A useful exercise once you have all your reverb sends set up is to start pulling the reverb return level down until you get to the point where the vocals just start to sound dissociated from the backing. You now have the region between this fader setting and your original aux return setting to play with in determining how dry you want your vocals to be. You may be surprised at how little reverb you actually need to keep the vocal sounding comfortable. Perhaps the best news is that there are no longer hard and fast rules on how much or what type of reverb to use. The trick is finding an approach that works with the track to achieve the desired artistic result.

Backing Vocals

For some tracks, the backing vocals are almost as important as the lead part, and it's always worth paying special attention to them in your mix. The process of recording and treating backing vocals isn't massively different from that of the main vocal part, although there are a few niceties to keep in mind. Firstly, if you're going to layer more than a couple of backing vocal lines, any breaths preceding the starts of phrases will start to stack up, so it might sound better to cut off the breaths from all but one or two parts. That should stop your backing-vocal entry sounding like a steam rally! Where you need to push the backing vocals to the 'rear' of the mix, you can add more reverb with less in the way of pre-delay, but this isn't always appropriate; as stated earlier, a layered vocal part often sounds adequately rich and textural with little or no added reverb. In that case, you can soften the highs using a high-cut EQ, which should sit the backing vocals behind the main vocal without using excessive reverb.

The timing of layered backing vocal parts is incredibly important in most modern productions. Obviously, spending time on rehearsals and ensuring that a good performance is captured are the best ways to make sure things are tight, but even then, edits are usually required. Backing-vocal timing can often be improved by viewing the relevant waveforms below the main vocal track and then moving errant phrases into line with the main vocal. Where a phrase starts at the right time but is of a different length to the main vocal, due to a slight difference in phrasing, your DAW's time-stretch facilities can often be brought to bear on individual phrases, words or even syllables to improve their alignment — whether that means making them longer or shorter. I don't like to go overboard with this kind of processing, though, as it can introduce audible artifacts, so don't always go by what you see on the screen: much better to rely on your ears when deciding whether or not the timing will benefit from tightening. Where you have to move two sections so that they overlap, make sure you use a short crossfade, which will help to disguise any edits.

Such manipulations may be too obvious on the more prominent lead-vocal part, but you can often take far more liberties with backing vocals, especially those comprising multiple layers, and those which sit underneath the lead (rather than appearing to 'respond' during gaps in the lead vocal.

In this screen shot, the timing of two backing vocal-parts has been tightened, by aligning them manually with the waveform of the lead vocal.

While it is just as common to use pitch correction on backing vocals as it is to deploy it on lead vocals, it pays to set this to be as relaxed as possible on layered parts: the faster the correction speed, the more likely it is that the various layers will be so close in pitch that you'll hear an artificial-sounding, chorus-like effect. A setting below half speed on a real-time pitch-corrector is usually OK, but try to use a slightly different speed setting on each layer. If you're using an off-line process, such as Melodyne or Cubase's VariAudio, the same rules apply. In other words, it's all too easy to overdo it.

Any of the fake double-tracking techniques that I described earlier in relation to lead vocals can be used to make a backing-vocal part sound more dense and complex — pitch-detuned copies, pitch-corrected copies, and even the application of a dedicated double-tracking plug-in, such as Antares Duo or Choir, can help. If you find that your layered vocals sound too weighty, you could try calming down the low end using a high-pass filter set at 150-200 Hz, which can help them stay in their rightful place behind the main vocal. Some overall compression and perhaps a little reverb is often beneficial for glueing the backing-vocal sound together, but do beware that if you overdo the compression you'll end up pulling the backing vocals too far forward, and they'll either sound out of place, or perhaps clash with the lead part.

Personally, I like to send the backing and lead vocals to a separate bus, as they'll usually need different overall processing, but very often I'll feed a send from each to the same parallel compressor, if parallel compression is appropriate, and they may also share send effects like reverb.

Practice Makes Perfect

That pretty much concludes this tour of contemporary vocal processing. As I said at the outset, we've tried to cover as much useful ground as possible, but there's much more to learn! If there is a key to creating professional-sounding vocals — other than tracking a great vocalist well — it is to put in plenty of hours of practice, and to research the sounds you like. You'll find a wealth of information on how different engineers approach vocals and other instruments in the regular Inside Track: Secrets Of The Mix Engineers articles in the pages of SOS, not to mention our Mix Rescues, in which we frequently use commercial tracks as a mix reference and describe precisely how and why an effect has been achieved.