Vocal Cords: Vocoder Basics

Heard any robot voices lately? Of course you have, because vocoded vocals are all over today’s charts. Vocoders have been used on hits before—such as Styx’s “Mr. Roboto” and Lipps Inc.’s “Funky Town”—but, today, they’re just as likely to be woven into the fabric of a song (Daft Punk, Air) as being applied as a novelty effect.

A vocoder has two inputs: Instrument (the “carrier” input) and Mic (the “modulation” input). As you talk into the mic, the vocoder analyzes the frequency bands where there’s energy, and opens up corresponding filters that process the carrier input. This impresses your speech characteristics onto a musical signal. Vocoders used to be expensive and rare, but now there are multiple vocoding options.

  • Standalone hardware units. You’ll find these mostly chez eBay. Sennheiser and Synton vocoders are prized, as is the famous Roland SVC-350. If you find one, buy it! For DIY fans, there’s PAIA Electronics’ 6710K 8-band vocoder kit (which I helped design).
  • Built-in vocoders. Several keyboards—such as the Korg R3 and Radias, Roland VP-550 and JP-8080, Alesis Ion, Novation Supernova, Access Virus C, and Nord Modular—include vocoders, because they already have the sounds needed to provide the carrier input.
  • Software plug-ins. Some programs—including Cubase, Logic, Sonar, and Reason—bundle in vocoders. However, a complication with plug-in vocoders is getting two inputs, because, until recently, the ability to sidechain a second input to provide the modulator (or carrier) was difficult to implement. Two common workarounds are to include a sound generator within the plug-in, and use the input for the mic (the approach taken by Waves’ Morphoder), or to insert the plug-in in an existing audio track, and use what’s on the track as the carrier (Sonar does this with Pentagon I; see Power App Alley in the 10/06 EQ).

Fundamental Sounds 

  • Talking instruments. To create convincing talking-instrument effects, use a carrier signal rich in harmonics, with a complex, sustained waveform. Remember, even though a vocoder is loaded with filters, if nothing is happening in the range of a given filter, then that filter will not affect the sound. Vocoding an instrument such as flute gives very poor results. A guitar will produce acceptable vocoding, but a distorted guitar or big string pad will work best. Synthesizers generate complex sounds that are excellent candidates for vocoding.
  • Choir effects. To obtain a convincing choir effect, call up a voice-like program (such as a pulse waveform with some low-pass filtering and moderate resonance, or sampled choirs) with a polyphonic keyboard, and use this for the carrier. Saying “la-la,” “ooooh,” “ahhh,” and similar sounds into the mic input, while playing fairly complex chords on the synthesizer, imparts these vocal characteristics to the keyboard sound. Adding a chorus unit to the overall output can give an even stronger choir effect.
  • Backup vocals. Having more than one singer in a song adds variety, but if you don’t have another singer at a session to create call-and-response type harmonies, a vocoder might be able to do the job. Use a similar setup to the one described above for choir effects, but instead of playing chords and saying “ooohs” and “ahhhhs” to create choirs, play simpler melody or harmony lines, and speak the words for the back-up vocal. Singing the words (instead of speaking them), and mixing in some of the original mic sound creates a richer effect.
  • Crowd Chants. Create the sound of a chanting crowd (think political rally) by using white noise as the carrier. This multiplies your voice into what sounds like dozens of voices. This technique also works for making nasty horror movie sounds, because the voice adds an organic quality, while the white noise contributes an otherworldly, ghostly component.

Added Tweaks

Some vocoders let you change the number of filters (bands) used for analysis. More filters (16 and above) give higher intelligibility, whereas fewer filters create a more impressionistic sound. Also, many speech components that contribute to intelligibility are in the upper midrange and high frequencies, yet few instruments have significant amounts of energy in these parts of the frequency spectrum. Some vocoders include a provision to inject white noise (a primary component of unpitched speech sounds) into the instrument signal to allow “s” and similar sounds to appear at the output. Different vocoders handle this situation in different ways.