BUT FIRST: A REALITY CHECK
If a singer is having pitch issues, don’t automatically reach for a fix: The problem may be a symptom of something deeper. For example, if the monitor level is too low and the singer can’t hear the vocals well enough to produce proper pitch, the vocals might be sung at too low a level as well, which leads to dynamics issues. Pitch problems might also mean a singer who’s tired, in which case the performance itself will suffer — not just the pitch. As with so many aspects of recording, check for the root of the problem and try to solve that, rather than put a band-aid on the symptoms.
And if you do need a fix, try punching first. I believe that processing such as pitch correction is a last resort to be used sparingly, unless you’re consciously using the process as an effect. No matter how little the pitch corrector messes with the sound — and these days, pitch correction algorithms are uncanny in their ability to fix problems unobtrusively — a natural vocal will always be, well, more natural.
Finally, just because you can fix something doesn’t mean you should. It’s like the early days of MIDI quantization . . . here’s a quick story. I had just fired up a Commodore-64 MIDI sequencer and called up Spencer Brewer, a superb piano player with a great “touch,” to check it out. He played into it, and as we looked at the data, he was shocked at how “bad” his timing was compared to the quantized “ideal.” So we quantized it, and his beautiful expressiveness disappeared faster than Kenny G at a Nine Inch Nails concert. The lesson: His timing was fine; it was the sequencer that didn’t understand music.
So it is with pitch correction. When you start using pitch correction, you’ll see that your voice has all kind of “pitch problems.” But despite the superb graphical tools present in today’s programs, listen with your ears, not your eyes. Fix only those notes that sound wrong. If a note’s a little off pitch, don’t worry about it. Quantizing pitch and rhythm too tightly can remove much of a vocal’s expressiveness.
HOW PITCH CORRECTION WORKS
Almost all pitch correction uses the same basic principle. The process begins by analyzing the input signal, then comparing it to a scale (usually user-definable). If there’s a difference, then the program pitch-shifts the original signal so it matches the ideal. However, how it applies this correction is subject to various user constraints. For example, you might want to have the input jump instantly to the new pitch, or slide to it over time instead.
There may also be various bells and whistles, such as being able to accentuate vibrato, “flatten” the natural vibrato and synthesize a new vibrato, change the formant (vocal characteristics, such as transforming male to female), and the like.
AND NOT JUST VOCALS...
Pitch correction is good for a lot more than vocals; any monophonic line is fair game (although the about-to-be-released Melodyne Studio 3 from Celemony handles polyphonic material as well). I’ve actually used Sonar’s V-Vocal feature more on bass than vocals to create slides and add “humanizing” elements to MIDI bass parts. Wind instruments and lead guitar can also benefit from vocal pitch correction tools. You can even apply processors designed for monophonic material to polyphonic material; the results will be unpredictable, but try pitch correcting a full drum set and see if it doesn’t have some potential for your latest post-punk industrial masterpiece.
Pitch correction can also be an effect (commonly known as the “Cher effect” because of the way correction was used in her hit “Do You Believe”). While the effect is now a cliché, extreme amounts can still sound cool sometimes.
BUT I WOULD RATHER SWIM IN SEWAGE THAN USE PITCH CORRECTION!
When Antares introduced the ATR-1 — the first pitch-correction-in-a-box device — in 1998, it not only fixed vocals, but brought new meaning to the term “machines don’t kill music, people do.” It was overapplied and misapplied to the point where pitch correction got a bad name. The irony was that people didn’t realize pitch correction could sound transparent, because when it was transparent, they didn’t know it was being used. Bottom line: Unless you’re applying fairly drastic fixes, no one will know you’re using correction — if you know what you’re doing.
DigiTech’s Vocalist line was one of the first families of commercially available products designed specifically for vocal pitch manipulation. The main way it worked with pitch was by synthesizing harmonies, either based on a preset scale or in reaction to MIDI note input.
Antares’ Auto-Tune was a step up, as it didn’t just shift pitch, but fixed pitch problems in near-real time. Their latest version, the ATR-1a (Figure 1), extends the correction range down to 25Hz (guess I’m not the only one who uses pitch correction for bass), and allows creation of custom scales to which pitches are referenced — as well as lets you correct only specific notes, like if the singer can’t hit high “C” quite right but gets everything else. You can also specify a melody with MIDI to which the pitch will “snap.”
Roland’s contribution to vocal processing was VariPhrase, although it was also intended for other instruments. It was innovative too, as it bridged the classic Antares approach with something more like the stretching options in Acid. It was ahead of its time, but its technology re-surfaced as part of Sonar 5’s V-Vocal processor, described later.
Also in hardware-land, TC-Helicon offers multiple pitch correction products: VoicePro, VoiceWorks, and VoiceOne 2.0 are studio processors, whereas VoiceLive is a floor processor optimized for live use. These also throw other functions into the equation, such as vocal modeling, doubling, reverb, and various effects.
THE SOFT SOLUTION
Just as Antares defined hardware-based pitch correction, Celemony’s Melodyne did the same thing for software. Its breakthough was to analyze incoming audio, and present it as graphical “blobs” (Figure 2) with pitch, rhythm, and amplitude information. You could actually “see” a melody with as much detail — actually, more so — than MIDI, and manipulate not just pitch but rhythm, formant, and dynamics as well.
Although Melodyne remains stand-alone at heart, the Melodyne bridge option is a plug-in that hooks into a host application, so Melodyne can run concurrently with the host. It’s also possible to ReWire Melodyne into a host (Figure 3), or ReWire clients into Melodyne’s mixer. This is where Melodyne’s ability to act as a stand-alone recorder comes in really handy. For example, if you ReWire Melodyne and Reason together, you can record vocals in Melodyne that run in parallel with Reason, but also do all the cool Melodyne editing tricks.
Although Melodyne Studio is expensive, there are two alternatives: Melodyne cre8 is quite similar but handles eight tracks instead of as many as your computer can handle, and goes up to 24-bit/48kHz instead of Melodyne Studio’s 32 bit/192kHz operation. There’s also Melodyne Uno, which handles only one track at a time — although often, that’s all you need.
Editing seems complex at first, because there are separate tools and sub-tools for editing amplitude, formant, pitch, and duration. But this is necessary because each process is quite different.
There’s also cool MIDI stuff: Control freaks can use MIDI to automate most Melodyne parameters, but even better, once you’ve detected audio,you can export it as MIDI data including pitch, dynamics, phrasing, and so on. You may need to touch things up, but if you ever wanted to “sing” a synth part, Melodyne is pretty effective.
DO THE RIGHT THING
Part of pitch correction is using the controls correctly. One of the main ones determines the speed with which the input pitch slides to the corrected pitch. Set this too fast, and the pitch “snaps” to the new pitch in a stepped, unnatural way (not that there’s necessarily anything wrong with that). Set it too slow, and there’s an almost equally unnatural slide effect that can negate vibrato and glissando. A lot of the bad attitude about pitch correction comes from misadjusting this parameter.
Speaking of vibrato, it’s often possible to either accentuate existing vibrato, which is a good thing if the singer’s vibrato is weak and you want to emphasize it, or “flatten” the pitch and add artificial vibrato, like the LFO does in a synthesizer. Unfortunately, some people think of artificial vibrato as what you add after you destroyed the original vibrato by setting the speed too slow. No! Try to preserve the existing vibrato, and choose accentuation over synthesized vibrato if you want the most natural sound.
On the other hand, if you have an angelic backup choir, you can get some interesting effects by flattening pitch and adding synthetic vibrato — it adds a kinda cool ethereal-meets-mechanical quality.
Another variable parameter sets the transition time between notes. Although you generally won’t want to add much, you can use transitions to create more of a “glide” effect between notes, sort of like synth portamento.
What may be the most important control sets the depth of the correction. At maximum, this clamps the vocal to the specified pitch. But no one sings that way; give a little slack and don’t overcorrect the pitch. After all, where would B. B. King be if he always bent held notes right up to pitch, instead of pausing, tantalizing, just before hitting the pitch?
MORE SOFTWARE SOLUTIONS
Antares made the move from hardware to software with their Auto-Tune 4 plug-in. You can use it the same basic way you’d use the box — plug ’er in and let ’er fix, with maybe a couple control tweaks — or a use a graphical mode where you can draw out pitch changes. Like most pitch correction software, the graphics are pretty intuitive and make it easy to see where the pitch deviates from the “ideal.”
TC offers the Intonator HS (Hybrid Shifting) realtime pitch correction plug-in (Figure 4) for the PowerCore platform, thus blending hardware and software. It uses the same basic intonation correction technology as TC-Helicon’s hardware products.
Waves recently introduced Tune (Figure 5), which aside from having the Waves pedigree, includes some unusual features like the ability to create custom scales and use non-Western tunings. It also supports ReWire; all edits are non-destructive, and saved as part of your host’s project. Another cool feature is the ability to export MIDI files for doubling instrument parts with synths or samplers, or even to do transcriptions.
What’s more, some host sequencers now include built-in pitch correction. Magix Samplitude includes an “elastic audio” window with an intuitive graphical interface (Figure 6); starting with Digital Performer 4.6, MOTU added a pitch automation function that allows drawing in pitch correction, directly within the host program.
Sonar 5 Producer Edition incorporates Roland’s VariPhrase techology in their V-Vocal plug-in (Figure 7), which offers separate tools for altering pitch, time, formant, and dynamics. The way it works is by converting standard audio clips into “V-Vocal Clips,” which run concurrently with your host. After tweaking, you can apply the edits to the audio clip, thus making them permanent.
SO... IS IT WORTH THE EFFORT?
Some people have commented to me that using pitch correction is a hassle, what with drawing in little changes and such. Hello?!? What they’re really saying is either the singer couldn’t sing (solution: new singer, not extensive pitch correction) or they were obsessing over the vocal just because they could. I recommend listening to a vocal, isolating the few notes that are a problem, tweaking those, and then forgetting you have pitch correction available.
It’s been said that the government that governs least, governs best . . . and so it is with pitch correction. Use it wisely, and you’ll be able to take that vocal that was perfect except for a few glitches, and ban those glitches forever — as well as do some really creative sound design tricks if you’re so inclined.