Vocoder

What it is and how it works

The vocoder (invented by Homer Dudley at Bell Labs, patents of 1937 and 1939, originally to compress speech in telecommunications) makes one sound "speak" with the voice of another. Two inputs:

• Modulator (usually the voice): split into many bands with band-pass filters; for each band an envelope follower measures how loud it is at every instant.
• Carrier (usually a harmonically rich synth, a chord, noise): passes through an identical filter bank; the gain of each band follows the envelope of the corresponding band of the modulator.

Result: the spectrum of the carrier takes on the spectral shape (formants, articulation) of the voice, but the pitch is that of the carrier. You play notes on the keyboard and the voice "sings" those notes, with the timbre of the synth. The original pitch of the voice is discarded. Many vocoders have an unvoiced channel (filtered noise) for sibilants and consonants (s, f, t), which the harmonic carrier otherwise doesn't reproduce.

Musical history: Robert Moog built one of the first solid-state musical vocoders in 1968; then the EMS Vocoder 2000, Roland VP-330, Sennheiser VSM201, Korg; used in pop and electronic music from Kraftwerk to Daft Punk.

Not to be confused with the talk box (the sound of an instrument sent through a tube into the musician's mouth and picked up by a microphone: Roger Troutman, Peter Frampton) or with Auto-Tune (pitch correction of the voice itself).

Parameters one by one

• Number of bands: 8–12 = robotic, hardly intelligible, vintage; 16–32 = good compromise; 40–100 = very intelligible, almost "synthetic voice". Vocodex goes up to 100 bands of variable width.
• Range / band distribution: where the bands sit (e.g. 100 Hz–8 kHz); denser bands in the 300 Hz–4 kHz zone improve the words.
• Attack / Release of the followers: attack 1–10 ms, release 20–80 ms. Long releases = more legato, "pad" sound; short = sharp articulation (but risk of roughness).
• Band width / Q: narrow bands = more "buzzy" and defined sound; wide = softer.
• Formant shift / band shift: shifts the carrier bands relative to the modulator: a "smaller" or "bigger" voice.
• Unvoiced / noise / sibilance: level of noise for consonants (20–40%).
• Carrier synth: Vocodex has a built-in carrier synthesizer; alternatively you use any instrument.
• Modulator / carrier level, dry voice mix: how much of the original voice to add (5–15% improves intelligibility).
• Hold / freeze: freezes the spectrum (static vocal pads).

Setup in FL Studio

• Vocodex (advanced vocoder, up to 100 bands, built-in carrier synthesizer, articulation envelopes and an internal Soundgoodizer) and Fruity Vocoder (simpler).
• Classic configuration: voice and synth arrive at the same mixer track as the vocoder, separated on the left and right channels (voice panned fully left, synth fully right, both with "Route to this track only"); the vocoder uses one channel as modulator and the other as carrier. Check in the "Setup Inputs" section of the Vocodex manual which channel is assigned to what and the alternatives (sidechain, internal synth).
• V Collection alternative: Vocoder V (emulation of the Moog vocoder) with internal or external carrier.
• Third-party: Waves OVox, iZotope VocalSynth 2, Arturia Vocoder V, TAL-Vocoder (free, inspired by 1980s vocoders), XILS 5000.

The right carrier

• Harmonically rich waves (sawtooth, narrow pulses, supersaw): more harmonics = more "lit" bands = more intelligibility. A pure sine doesn't work.
• Wide chords (three to five notes) for vocoder choirs; a single note for a robotic lead.
• Add a little noise to the carrier or use the unvoiced function for the consonants.
• Carrier already compressed and stable: the level must come from the modulator.
• Carrier range consistent with the voice (centered between C3 and C5 for normal voices).

Preparing the modulator

• Dry voice (no reverb), compressed (GR 6–10 dB) and with emphasized presence (+3/+6 dB at 2–5 kHz): more articulation.
• HP 100–150 Hz.
• You can also vocode drums and loops (carrier = pad): "talking" rhythms.

When to use it and on what

• Robotic lead (electro, French house, 1980s funk, synthpop).
• Vocoder choirs as filler in choruses (chords + lead voice).
• Vocal pads and ambient textures.
• Effects on intros, bridges, breakdowns, ad-libs.
• Vocoded drums on a pad (rhythmic effect).

Starting settings

Mixing the vocoder

• EQ: HP 120–200 Hz (the vocoder often brings a lot of low-mids from the carrier); −2/−3 dB at 300–500 Hz; presence +2 dB at 2–4 kHz.
• Compression after the vocoder (the bands vary a lot): GR 3–6 dB.
• Chorus or ensemble for width; plate or hall reverb on a send.
• Level: like a secondary vocal if it does choirs, like a lead if it carries the melody.

By genre

• French house, nu-disco, funk, electro: vocoder leads and choirs as a signature.
• Synthpop, synthwave: vocoder choirs and pads.
• Hip hop and R&B: vocoder choirs and talk box (G-funk style) on hooks.
• Ambient: vocal pads.

Common mistakes

• Carrier poor in harmonics: muted, incomprehensible sound.
• Voice with reverb or noise as the modulator: everything opens the bands, articulation is lost.
• Too many low-mids from the carrier: muddiness with pads and guitars.
• Unstable carrier level: the vocoder "pumps".
• Confusing vocoder and Auto-Tune: if you want the voice in tune but human, the vocoder is not the right tool.

How to hear it

• Try the same modulator with 8, 24 and 100 bands: hear how intelligibility grows.
• Remove the unvoiced: the "s" sounds disappear; put it back until they sound natural again.