How a Vocoder Works
A vocoder needs two signals. The modulator, usually a voice, supplies the words, and the carrier, usually a synth, supplies the pitch and the tone. Neither is heard directly: what comes out is the carrier, reshaped moment by moment by the modulator.
The reshaping happens in bands. The vocoder splits the modulator into a row of frequency bands and puts an envelope follower on each, so it knows how loud every band of the voice is at every instant. It splits the carrier with a matching set of filters and sets the level of each carrier band from its partner in the voice.
The result prints the voice’s formants, the resonant peaks that make an “ah” differ from an “ee”, onto the synth, while the voice’s own pitch is thrown away. That is why a vocoder is played rather than applied: hold a chord and speak, and the chord talks. It is also why it differs from pitch correction, which keeps the singer’s voice and moves its pitch, so a vocal tuning plugin driven hard gives a stepped robotic voice while a vocoder replaces the voice with a synth.
Bands and Intelligibility
The number of bands sets how finely the voice’s spectrum is copied. Few bands give a smeared, characterful result where vowels survive and words blur, and more bands give a sharper copy that reads as speech. Vocoder plugins vary widely here: Softube’s Vocoder offers 4, 8, 12, 16 or 20 bands, while MeldaProduction’s MVocoder goes up to 100.
More is not automatically better. A low band count is often the sound wanted for a hook or a pad, and very high counts can sound like a filtered voice rather than a talking synth. The useful question is whether the count can be changed, so one plugin covers both.
Consonants are the other half of intelligibility. An S, a T or an F is noise rather than a pitched sound, and a carrier holding a chord has little energy where those consonants live, so they vanish and the words go soft. Better vocoders pass noise or the voice’s own high frequencies through when a consonant arrives, which is the job of the Unvoiced section in Softube’s Vocoder.
The Carrier Decides the Sound
Because the carrier supplies all of the tone, it matters as much as the vocoder. A sawtooth or pulse wave is rich in harmonics, so every band has something to shape and the result is bright and intelligible, while a sine or a soft pad leaves bands empty. Noise as a carrier gives a whisper, and a slow chord on a warm pad gives the classic synth choir, which is why an analog-style synth from our analog synth roundup is the usual starting point.
The carrier does not have to be a synth at all. A string section or a drone works, and so does swapping the usual roles: a drum loop as the modulator imposes its rhythm onto a sustained pad, so the pad pulses in time. Softube’s Vocoder builds in a six-voice carrier synth played from MIDI, MVocoder takes its carrier from a sidechain input, and iZotope’s VocalSynth 2 offers both.
Classic and Modern Uses
The vocoder began at Bell Labs in the 1930s as a way to compress speech for telephone lines. It became a musical instrument in the 1970s, where it gave electronic pop its talking synths and science fiction its robot voices, and the synth choir from that era is still a stock sound.
In media work the modern uses are mostly texture and character. Robot, AI and creature dialogue is the obvious one, built from an actor’s line and a carrier chosen to suit the character. Less obvious is the rhythmic bed: a sustained string or synth pad vocoded by a percussion loop locks to the cue’s tempo without a single automation lane.
VocalSynth 2 puts a vocoder in one window with four other voice engines, a talkbox among them, which helps when the brief is to make a voice synthetic and the method is open.
What to Check Before Buying
Band count, and whether you can change it. A fixed low count is one sound. An adjustable count is an instrument.
Where the carrier comes from. A built-in synth is the faster route. A sidechain input lets you use your own synths and orchestral patches, so check how your DAW routes sidechain audio into an effect before relying on it.
Consonant handling. An unvoiced, sibilance or noise control decides whether the words survive.
Format. The best vocoder VST and the best vocoder software are the same tools under two names. VST is one plugin format among several, alongside AU and AAX, so check that a vocoder runs in your DAW rather than which acronym it carries.
Getting a Vocoded Part to Read
Prepare the modulator like a vocal you want heard. Compress it firmly so every syllable drives the envelope followers at a similar level, and a gentle presence lift before the vocoder helps the consonant bands respond. Blending in a little of the dry voice, high-passed so only the consonants come through, is the standard way to recover the words without losing the effect.
