Perfect Pitch Correction
Working with pitch is an essential part of modern music production, whether you’re making subtle tweaks or going for the full robot effect. Hollin Jones hits the right notes.
For most of the 20th century, correcting the pitch of recorded audio without also altering its speed was a decidedly tricky process. Its analogue nature meant that it had to be played mechanically, usually from a tape. Anyone who has played around with the speed of tape or vinyl knows that running it slower will lower the pitch and running faster will raise the pitch, but this poses a problem if you’re trying to correct out-of-tune parts that need to remain at a specific speed. Engineers therefore conceived necessary tricks to get around this problem, such as slightly altering the tape speed until a part matched the pitch of an untuneable instrument that was sharp or flat, or playing with sync through different tape heads to match up one part with another. But not only did this require ingenuity, you also needed access to the right kit. Such tricks with tape have been used extensively to create harmonies from individual vocal tracks, but this relied somewhat on the original take being in-tune. However, there are all kinds of ways in which you can creatively layer harmonic variations – provided your source material is on the money.
Digital world
When digital audio came into being it was initially employed mostly in samplers, which were then highly expensive and available only to the few. Pitch-shifting was easy, but pitch correction within audio clips, at least in real time, was still difficult. The problem was that algorithms had not been devised that could deal with analysing and correcting audio on-the-fly – and the processors used in samplers and computers weren’t powerful enough to do it anyway. It was possible to correct pitch if you had digital audio files, but this generally involved working on very small sections of sound simultaneously and rendering or bouncing down to make the changes permanent.
Changing times
Quite apart from the technical difficulties involved in pitch correction in those days before software, it’s worth thinking about why we need to do it at all. Most ‘real’ instruments can be tuned – so they can also be out of tune. An engineer might therefore have to deal with a brass session, for example, that was slightly out, or try to fit a piano part into an arrangement despite it being fractionally out of tune.
It’s also possible that tuning was simply less of an issue in days gone by due to the way in which people worked. The fact that recordings used to take place far more frequently in a single room containing musicians – today they are more likely to be recorded at different locations then layered together – put an emphasis on getting it right first time. There was a much higher chance that all of the instruments would be in tune with each other. And as controversial as it might sound, it’s also quite feasible that singers were, on the whole, better able to hit and hold a note than they are today. That’s partly because far fewer people got the chance to sing in a professional studio, so statistically they were likely to be better singers. You got fewer takes than you do today (prior to tape you would be lucky to get more than a couple) so you had to be good.
What’s more, the digital age has brought with it an expectation of perfection that didn’t really exist beforehand. This is part of a wider grumble that some people have about modern music and takes in auto-tuning, the Loudness Wars and ‘in-the-box’ production. Listen to a record from 30 years ago and the chances are that it won’t be absolutely perfect: there will be the odd mic pop, some bleed between tracks and some notes that aren’t precisely on the money. But no one criticised such records for such things because they were not perceived as flaws – they were just a natural part of the performance. Indeed, they actually give a character to a record that some people feel is altogether lacking in modern production.
In with the new
More recently, pitch correction has become far easier and much more accurate thanks to the development of algorithms and plug-ins that can correct in real time. As is so often the case, these started life as very high-end products, but the technology has now filtered down to a more easily accessible level (Cubase, Logic and Record, among other DAWs, all ship with advanced pitch-correction tools). Other developers have been hard at work as well, with Bias’ PitchCraft (£122, www.bias-inc.com), Antares’ Auto-Tune (£310/£499, www.antarestech.com) and Celemony’s Melodyne (£89–£305, www.celemony.com) all offering their own unique take on fixing pitch. Most of them work best on monophonic sources such as vocals, since polyphonic audio – in which more than one pitch occurs at the same time – is much harder to work with. Nonetheless, this has recently become possible: Melodyne Editor has something called Direct Note Access, which allows you to ‘reach’ inside chords and alter individual notes, even when they occur at the same time as other sounds within a single audio clip. Auto-Tune 7 also enables you to manipulate timing and pitch within an audio file.
As time goes on, pitch-correctors are becoming even more powerful and accurate, incorporating, for example, tools that enable you to extract the pitch from audio and create MIDI notes from it. Cubase’s VariAudio system is particularly adept at this, even translating micropitch – tiny variations in pitch as you slide from one note to another – into MIDI CC data such as pitch bend. This enables you to extract MIDI from any monophonic audio part, then use a virtual instrument to re-create or back up the same melody precisely.
Less or more
Before we go on, there’s a distinction to be made about the degree of pitch correction we’re talking about. What you might call ‘subtle’ pitch correction is used extensively on all kinds of high-profile albums these days. Arguably, this is because record labels feel that their acts should be ‘perfect’, especially when so much money has been invested in them. In fact, it’s so prevalent that you almost don’t notice it, and it’s easy to assume that all voices in pop are flawless when they aren’t.
The flipside of this is what is now generically referred to as ‘auto-tuning’: the deliberate use (or even misuse, depending on your opinion) of extreme pitch correction to produce the robotic ‘stepping’ effect made famous by Cher on her single Believe and later T-Pain, Kanye West, the Black Eyed Peas and countless house tracks. It became so popular that Jay-Z wrote the song Death Of Auto Tune, on which he deliberately sang out of key.
Auto-tuning is essentially very strong pitch correction that straitjackets the signal into strict notes, not allowing any variation or glides between them, creating ‘steps’ rather than the natural sound of a voice. Its effect is comparable to vocoding, but the process is different. Auto-tuning modifies an audio signal; vocoding uses a vocal as an input to synthesise new sounds based on keys that you press. Vocoding does create exact pitches based on vocal input, but with the important distinction that the sound is synthesised, not pitch-corrected. This idea of using keys has made it into some pitch-correction plug-ins, enabling you to specify MIDI notes and tell the software which note you’re trying to correct to.
The question of whether to tune (and if so by how much) is a difficult one. It comes down to personal preference, or the wishes of the client you may be working for. There are two main forms of pitch correction, which could be termed ‘real-time’ and ‘grid-based’. Which one you use will depend on the plug-ins you own – some do one or the other and a few do both – as well as the material you’re working with. It’s a given that before you apply any pitch correction you should have reasonable grounds to believe that there is something not quite right about the tuning of the sound. This doesn’t apply to the robotic auto-tune effect, since it is so mechanical. But for regular vocals, identify the problem before you apply a solution.
Pitch perfect
Real-time correction takes a live stream of audio from an audio track or, in the case of a live show, a mixer channel. The plug-in analyses the incoming audio in real time and applies correction based on the settings you have made regarding key, accuracy and formants. Real-time correction works best on audio that is only a little bit out of tune – if it’s any worse than that the software can struggle to determine which note it is supposed to be pulling the sound towards and get it wrong. With some careful setting up you can get around this problem, creating a special map of notes and settings tailored exactly to a particular passage or song, so that there are no unexpected bum notes. This can take some time and necessitate repeated playings, but it’s quite feasible.
It’s rare that simply applying pitch correction as an insert effect will work from the get-go; it’s almost always the case that some tweaking is needed to get it working properly on your material. This is less true of auto-tuning, as it is much more of a blunt instrument, untroubled by the subtleties involved in tuning a normal vocal part.
On the grid
Pitch-correction tools that use grid systems (such as those found in Melodyne, Cubase’s VariAudio system and Antares’ Auto-Tune) are better at dealing with specific parts of an audio clip since they are able to break down the frequencies and amplitudes into a visual grid, from where they can be edited with the mouse. Obviously, if you tried this in real time the display would move too quickly to be of much use. Such tools are capable of subtle or extreme pitch manipulation, but are especially good on parts of a take where something is more out of place than the rest of the track. You might, for example, identify a couple of bad notes in an otherwise great take; in this scenario, applying a real-time pitch corrector might be overkill, or even mess with the in-tune audio.
Of course, you also have the option of using automation to turn pitch correction on and off as an insert effect at certain points, which is an ideal way to employ real-time correction at certain points on a track but switch it off at others. You could even have several copies of the same vocal track running, each with a pitch-correction plug-in applied (with different settings for different parts of the track), selectively muting them as necessary using automation. This approach could be used to correct an out-of-tune vocal in a track that changed key mid-way through: one instance of a plug-in would deal with the vocal in one key, while a second addressed the other. Alternatively, you could automate between settings or presets on your plug-in so that the same plug-in on a single track behaves differently at different points.
Grid-based pitch correctors give you more hands-on control because the audio can be isolated, looped and edited in your own time rather than playing back as part of the song. That said, this method does encourage you to fiddle endlessly, making tiny adjustments, so bear that in mind. You have to wonder at what point a vocal is so bad that it might be better simply to re-record it, although this is often not possible due to time/cash constraints.
Present and correct
When setting up pitch correction your procedures will depend on the type you are using – real-time or grid – but some of the same principles will apply in either case. You will need to know what key you’re working in and whether it’s minor, major, chromatic or some more exotic variation on a scale. Before you get into really advanced settings, the main thing is to get the notes right, so in the case of a real-time correction plug-in, loop up the part that requires correcting and identify which notes are being altered. You can use the mouse or a MIDI keyboard to determine which notes are allowed and which are not. If a note is being corrected to C #, for example, but it should be going to D, you can deactivate C # so that it won’t be heard. Of course, there may be another point in the song where C # does need to be heard, so you need to be careful and work on sections of audio at a time.
Pitch correction isn’t a ‘fire and forget’ kind of process, and you often end up automating a plug-in on and off, switching between settings or bouncing down sections of corrected audio a little at a time so that the track sounds right for its entire duration. Plug-ins almost always come with some general presets for different kinds of material, so you might be able to choose, for example, male or female, high or low or instrument/bass, which will help the plug-in to select a suitable algorithm and get you started.
MIDI control of pitch correction is a relatively new idea and features in Auto-Tune 7 as well as Cubase. In addition to specifying notes to use for correction, you can route MIDI either from a track or a keyboard to the plug-in and base the correction on the notes specified in the MIDI data. This also works with timing, so it’s an interesting way of messing with existing audio parts and creating something completely new.
Once your basic notes have been established you can move on to the more advanced controls. Pitch-correcting vocals and instruments can often iron out the natural vibrato of the sound and make it slightly unnatural. Today’s powerful plug-ins are able to deal with this by compensating for vibrato during correction or reintroducing it afterwards, having controls that you can edit. Auto-Tune 7, for example, has Humanize, Formant and Create Vibrato controls as well as things you would expect, such as the ability to precisely control the speed and character of the basic pitch correction that is being applied.
Auto-Tune 7 also has a graphical editor, as does Melodyne Editor edition. In these views you can work with individual notes and sounds, editing not just in steps like tones and semitones, but far smaller increments, making subtle pitch adjustments instead of major ones. Melodyne Editor provides a really neat way of working with mono or polyphonic audio, offering a simple graphical view of the notes and their positions and amplitudes. You can then easily edit the pitch, formant, amplitude and timing of each one without causing unwanted side effects and without altering the rest of the clip. By drawing around several notes you can edit them simultaneously.
You can fix it
The extent to which you use pitch correction in your productions depends very much on your personal preference. There’s certainly an argument to be made that people should be able to sing or play properly and that the odd minor inconsistency adds character to a song rather than detracting from it. On the other hand, there’s nothing worse than a sharp or flat note ruining an otherwise good take. If you’re working commercially, the decision may be taken out of your hands: if a client wants a recording perfectly tuned after the fact, you’ll probably have to do it – and it’s now very common.
Whether you use it a little or a lot, it’s good to know that pitch correction is possible and that the technology is more powerful today than ever. MTM
This feature first appeared in Music Tech Magazine issue 99
Filed under
General Features,
Features
Sign in to download this article
New users, please register here
See also...
 MTM 107 |
25 Tips for Mix Processing |
MTM 107 |
Control for Live Performance |
|
 |
Music Tech Focus - Synthesis 2012 |
MTM 106 |
25 Pro Tips for Kick-starting Composition |
|
MTM 106 |
Contemporary Mastering |
MTM 105 |
Re-creating the Sound of Old-skool Dance |
|
MTM 105 |
25 Tips for Live Sound |
MTM 105 |
Mixing Tips from the Industry Pros |
