Production – Repairing Pitch-Shift Artifacts

CELEMONY MELODYNE is an incredibly useful plug-in for arranging background vocals.
Image placeholder title

Fig. 1. The equalizer plug-in in ozone 5 advanced applies matching EQ to a pitch-shifted vocal. Note the severe EQ (the off-the-scale red line in the top section of the GUI) needed to correct comb filtering and roll-offs caused by pitch shifting.

Image placeholder title

CELEMONY MELODYNE is an incredibly useful plug-in for arranging background vocals. Working with a copy of the lead vocal, you simply drag each melody note lower or higher in Melodyne’s GUI to hear how the resulting pitch will sound as a harmony part in combination with the original lead vocal track. In this way, you can compose contrapuntal (non-parallel) harmonies and immediately hear them without singing or recording a single note. Pitch-shift intervals greater than a third will typically incur ugly processing artifacts—muddiness and a pinched, phasey sound—but who cares? After all, these are just throwaway tracks used to test your arrangement.

Or not. My dark secret: I’ll often use Melodyne-generated BVs—alone or in combination with real singers—in my final mixes. The impeccably tight phrasing and familial diction cloned from the lead vocal track are often impossible for real performers to trump. The proviso is I have to make those raggedy pitch-shifted tracks sound smooth enough to use.

In this article, I’ll share tips for diminishing the injury caused by large-interval pitch-shifting. There’s no miracle cure. But using these pointers, you can expect to recover roughly 25% of the original audio fidelity— good enough that, blended with the lead vocal and mixed into a fairly dense ensemble recording, nobody will be able to tell the BVs were not sung by live performers. Our first-aid treatment begins with restoring clarity.

Remove Masking Distortion Bounce each harmony part (with pitch-shifting applied) to a separate track so you can process it independently. Instantiate a broad-spectrum noise-removal plug-in on the track. The goal here isn’t so much to remove hiss as it is to sponge out any distortion artifacts having characteristics similar to noise. iZotope Denoiser, part of the company’s superb RX2 plug-in bundle, works great for this purpose. That said, the improvement in clarity will be subtle, as there will remain yet another type of poison to bleed out.

Fill in the Gaps One of the most toxic side effects of pitch-shifting tracks is deep comb filtering. One way to add back missing frequencies in the spectral notches is to process the rendered track with a harmonic exciter. I like to use the component Exciter plug-in in iZotope Ozone 5 Advanced for this application, inserted after Denoiser. (The exciter module in the all-in-one Ozone 5 plugin works equally well for this application.) Exciter’s multiband processing allows me to add harmonics to only the midrange band, where the damage to pitch-shifted vocals is typically most apparent. Exciting the track fills in the spectral gaps a bit, resulting in a sound that’s a bit fuller and more natural.

Match the EQ The Matching EQ mode in Ozone 5 Advanced’s Equalizer plug-in—or the equalizer module for the integrated Ozone 5— can be used to reverse comb filtering and other timbral damage to a pitch-shifted track. The crux is to use Matching EQ to reproduce the lead vocal’s native spectrum in its pitch-shifted spin-off. Instantiate Equalizer on the lead vocal track. In the Snapshots tab, capture the spectrum for one phrase of the vocal. Click on the Save Set button in the plug-in. Open another instance of Equalizer on the pitch-shifted track, and click on the Load button in the Snapshots tab. Choose the lead vocal’s spectral snapshot as the reference, and apply it to the pitch-shifted track in Matching EQ mode. Adjust the Amount and Smooth sliders for the most pleasing timbre; for tracks shifted down in pitch, I find respective settings of 50 and 0.3 work well. You’ll get the best results if you capture the unique spectrum for each vocal phrase in turn and apply matching EQ line by line.

If your equalizer plug-in can’t execute matching EQ but provides a spectrogram, you can manually fashion an inverse EQ curve to reverse comb filtering. In your DAW, loop a short phrase in the pitch-shifted vocal track. Looking at the peaks and notches in your equalizer’s spectrogram as the phrase plays back, fashion a set of bell-curve filters that will together create an inverse EQ curve to flatten the response. While you’re at it, correct any low- and high-frequency roll-offs to taste with EQ, too.

The peaks and notches will change from one vocal phrase to the next, so you’ll have to work in very short sections and readjust your filter parameters as you work on each phrase in turn. Once you’re satisfied the EQ is sounding as good as you can make it for the current vocal phrase, render all processing (de-noising, harmonic excitation, and EQ). The result will sound far from perfect in isolation, but significantly better than the original pitch-shifted track.

Use All Three Tools Used alone, any one of the processing techniques I’ve detailed will yield only subtle effect. Used together, however, they can make a significant improvement in the fidelity of pitch-shifted vocals. But don’t stop there. Double one or more of the pitch-shifted vocal parts with live-performance overdubs, and the composite effect can sound truly amazing.

Michael Cooper is the owner of Michael Cooper Recording in Sisters, Oregon (, and a contributing editor for Mix magazine.