A proper analysis/synthesis vocoder, not a pitch-shift trick. Your voice is split into as many as sixty-four bands and used to shape a carrier, so what comes out keeps your words and takes on the carrier's tone. Nine presets, a modulation matrix, seven effects and a recorder.
Your voice is the modulator and the synth is the carrier. A filter bank measures the energy in each band of your voice and imposes that shape on the carrier - which is exactly what a hardware vocoder does, and why the band count changes the character so much.
Band count, bandwidth, formant and carrier are all on the surface - which is the difference between a preset toy and an instrument.
Low band counts are coarse and robotic - the classic sound. High counts stay intelligible and almost natural. Sweeping the count while talking is the fastest way to understand what a vocoder actually does, and it is a single slider here.
Narrow bands ring and separate; wide bands blur into each other. Combined with resonance, this is what decides whether the result sounds like a talk box, a choir or a broken radio.
Move the formants and the apparent size of the speaker changes - up for small and bright, down for monstrous. The carrier itself runs from 20 Hz to 4 kHz, so the tone underneath your words is yours to choose.
Two LFOs, two envelopes and velocity, routable to the parameters that matter. Put a slow LFO on formant shift and a static line starts to move; put an envelope on band depth and consonants punch through.
Reverb, delay, chorus, phaser, distortion, bitcrusher and a compressor. Chorus and reverb turn a single voice into a choir; bitcrusher and distortion are how you get the cyberpunk end of the range.
Gain, a noise gate and sensitivity in front of the analysis stage, with a live spectrum on both the input and the output. A vocoder is only as good as what you feed it, and this is where most bad results get fixed.
Load one, then move the band count - it is the control that changes everything.
Low band count, narrow bandwidth, steady carrier. The sound everyone means when they say vocoder, and the best place to check your input level.
High band count with chorus and reverb wide open. Layered, soft and surprisingly convincing on sustained vowels.
Formants shifted down hard, carrier low, distortion engaged. Enormous, and completely unusable for anything except exactly what it is for.
Mid-band emphasis, bitcrusher on, everything outside the speech range gone. Retro radio, intercom, bad phone line - all the same patch.
The vocoder needs a live input, so the browser will ask for microphone permission. Nothing is uploaded - the audio is processed in the page.
Raise the gain until the input spectrum is moving properly, then bring the noise gate up until the room disappears between words. Almost every disappointing vocoder result is an input problem.
Robot is the one to start on because it makes mistakes obvious. Speak steadily and clearly - vocoders reward diction more than volume.
Sweep it from 4 to 64 while talking. Somewhere in there is the balance between "clearly a machine" and "clearly a word", and that point is different for every voice.
| Type | Analysis/synthesis vocoder and vocal processor |
|---|---|
| Bands | 4 to 64, selectable - default 32 |
| Band width | 0.1 to 2.0 |
| Carrier | 20 Hz to 4 kHz, with formant shift and resonance |
| Input stage | Input gain, noise gate, sensitivity, compressor |
| Modulation | Two LFOs, two envelopes and velocity, routed through a modulation matrix |
| Effects | Reverb, delay up to 1000 ms, chorus, phaser, distortion, bitcrusher (1–16 bit), compressor |
| Presets | Robot, alien, choir, monster, whisper, cyberpunk, telephone, vintage, underwater |
| Metering | Live input analysis, output spectrum, waveform display |
| Recording | MediaRecorder capture with WAV export; presets save and load locally |
| Requirements | Any modern browser with Web Audio and microphone access. |
A vocoder is not a pitch shifter and it is not an autotuner. It splits one signal - your voice - into frequency bands, measures how much energy is in each, and uses that measurement to shape a second signal. Vocoder Talk Box implements exactly that on the Web Audio API, with the band count, the bandwidth, the carrier and the formants all exposed rather than hidden behind a preset list.
Four bands is barely intelligible and unmistakably robotic. Sixty-four is close to natural speech wearing someone else's tone. Everything interesting happens between those two, and because it is a single slider you can find your voice's sweet spot in about ten seconds - then move it deliberately as an effect.
Input gain, a noise gate and a compressor sit in front of the analysis stage, both spectrums are visible while you work, and a recorder captures the output as a WAV. Presets save locally, so the patch you spent twenty minutes on is still there tomorrow.
The vocoder shares the studio audio graph with the synths, drum machines and samplers, so the carrier can be another instrument and the output can feed the rest of the chain. One subscription covers every app, and a 7-day free trial covers the deciding.
A voice changer usually shifts pitch or formants and leaves your voice as the source. A vocoder uses your voice only as a shape and applies it to a second sound - the carrier. That is why the result keeps your words but takes on the carrier's tone entirely.
Between 4 and 64, and it depends on what you want. Low counts sound coarse and mechanical, high counts stay articulate. Thirty-two is the default because it is intelligible while still obviously processed; sweep the slider while talking and pick your own.
Yes - the vocoder needs a live input to analyse. Your browser will ask for microphone permission the first time. The audio is processed inside the page and is not uploaded anywhere.
Almost always the input stage. Raise the input gain until the analysis spectrum moves properly, then bring up the noise gate so the room falls silent between words. Speaking clearly matters more than speaking loudly - vocoders reward diction.
Yes. The recorder captures the processed output and exports it as a WAV file, and your own presets are saved locally so you can load them again in a later session.
It is part of the GOONS Studio subscription (€9.90/month), which unlocks every app in the catalog, the 5000-sample library and cloud projects. New accounts get a 7-day free trial with full access and no card required.
Vocoder Talk Box comes with the GOONS Studio subscription at €9.90/month, alongside every other app in the catalog. Start with the 7-day free trial - no credit card, cancel whenever.
Already subscribed? Launch Vocoder Talk Box →