Before we released VocalSynth 2, we shared “3 Ways to Warp Vocals with VocalSynth.” The techniques ranged from ethereal to dark to highly synthesized. With VocalSynth 2 out now, we’re taking things a step further.
With the following experiments, we play around with layering VocalSynth 2 modules. Some vocal samples have two, three, even four VocalSynth 2 modules per track.
Multiply your pitch correction
Typically, when we hear pitch correction, it’s probably one instance of it. But with VocalSynth 2, we decided to play around with multiple layers of differently pitch-corrected vocals.
For Track 1, we set the vocal’s pitch to Minor A, and optimized it at Low. (Keep in mind that the vocal isn’t in the lower register, but we just wanted to see what happened if we played with this setting, and it seems to give the vocal a smoother flow.) The pitch correction is also at maximum strength and speed, but the Mix is set about halfway between Wet and Dry.
Track 2’s vocal is also in Minor A, and optimized at Low. This time, however, Strength is set to 69 and Speed is at 24. The Mix is 100% wet. When combined, the slight variances in the pitch correction create a strange electronic quality that serves as an interesting counterpoint to the original dry mix.
Formant shift vocals into oblivion
This experiment started with a question: What would happen if several layers of VocalSynth 2 modules were stacked on a single vocal track, each at different formant? Well, quite a lot actually.
But before we get into the approach and results, what exactly is formant shifting? Well, it’s easy to understand when comparing it to pitch shifting. In pitch shifting, whether it’s being done on a voice or instrument, the key of the notes are either shifted up or down. With formant shifting, on the other hand, the voice or instrument can go higher or lower without changing keys. So with the sample below, we took a vocal sample, and applied layers of VocalSynth 2 modules, each shifted to a different formant.
Track 1 is set to the Polyvox module with no pitch correction, and features a deep -12 formant. Track 2’s Polyvox module is set to a -6 formant. For Track 3, we set Polyvox to +6 formant, while Track 4 features +12 formant. We added a fifth track, which is a dry version of the vocal sample, so that we can still hear the singer’s original phonetic vocal sound.
In the Wet recording below, each track’s formant is introduced track-by-track, alongside it’s preceding track, until all play at once. Thus you hear Track 1, then Track 1+2, then Track 1+2+3, and so on. The overall effect is quite synthetic and rich in formants, like something you might hear on a UK garage or dubstep record.
Turn your voice into a strange computer
Back in 2001, Gallic duo Air released 10,000 Hz Legend, a masterful record of shape-shifting, both from track to track and within the same song. On the track “Wonder Milky Bitch,” Nicolas Godin and Jean-Benoit Dunckel craft one of the strangest lounge pop tunes to ever grace a pair of speakers.
It’s as if Spaghetti Western composer Ennio Morricone collaborated with Lee Hazlewood on soundtracking a science fiction film from the psychedelic 60s. The song features a spoken word narrative, along with whistling, but something is just off about the vocals. Air might well have used a Boss VT-1 Vocal Transformer on this song, and the result is that it sounds like a computer simulating Lee Hazlewood’s baritone without sounding too computerized.
This is what we’re after below. We didn’t want the vocal sample to sound too computerized, like a Daft Punk track, or even some of Air’s other work. We wanted a more subtle approach to computerized vocals.
For our clip below, we multitracked the vocal across two different tracks, and used the Compuvox and Polyvox modules for both. On Track 1, we set the Compuvox module to the Hushed Murmur preset, then used Polyvox to formant shift the vocal to +2, and also added some filtering and chorus effects. We then used Compuvox’s Quirky Creature on Track 2’s vocal, and shifted Polyvox’s formant to -2, with no effects.
The results are pretty much what we were after from the beginning. The vocals aren’t too computerized. Instead, there is a nice blend between what sound like natural vocals and simulated ones, like on Air’s “Wonder Milky Bitch.”