Vocoder
Channel vocoder: a modulator on Key imposes its band envelopes on the carrier through 8 to 32 band-pass filters, with formant shift and sibilance.
- Effect
- Since 0.2.0
- vocoder
- voice
- robot
- formant
- talkbox
- filterbank
- sidechain
Make any sound speak: a voice, drum or noise on Key shapes the spectrum of a synth or chord on Audio, band by band.
What it does#
The vocoder splits both signals into the same set of bands, spaced evenly in pitch from Low band to High band. For each band it follows the level of the modulator (the key input) and sets the level of that band of the carrier (the audio input) to match. The carrier supplies the pitch and the tone; the modulator supplies the moving shape of vowels and consonants. Each band is a pair of band-pass filters, and the envelopes rise with Attack and fall with Release. There is no added latency.
With nothing on Key, the vocoder uses its own input as the modulator, which turns it into a smooth band-envelope filter. A one-lane modulator drives every lane of a wider carrier. The level is set so a broadband modulator at about −12 dBFS RMS leaves a broadband carrier near its own level, whatever the band count and width.
When to use it#
Use it for robot voices, talking synths and pads, drums that "speak" through a chord, or rhythmic gating of a sustained sound by another part. Put the sound with the pitch on Audio and the sound with the rhythm or words on Key. For one moving resonance rather than a whole bank, use the Filter.
Ports#
| Port | Direction | Signal | Lanes |
|---|---|---|---|
audio | Input | Audio | Inferred from the cable |
key | Input | Audio | Inferred from the cable |
audio | Output | Audio | Inferred from the cable |
Parameters#
| Parameter | Label | Range | Default | Unit | Scale | Notes |
|---|---|---|---|---|---|---|
bands | Bands | 0 to 4 | 2 | linear | Number of bands, spaced evenly in pitch from Low band to High band. More bands make speech clearer. Choices: 0 8, 1 12, 2 16, 3 24, 4 32. | |
low_hz | Low band | 20 to 2000 | 100 | Hz | log | Centre of the lowest band. |
high_hz | High band | 500 to 16000 | 8000 | Hz | log | Centre of the highest band, kept at least an octave above Low band and below 0.45 times the sample rate. |
width | Band width | 0.5 to 2 | 1 | × | log | Band width relative to the band spacing: 1 makes neighbouring bands cross at −3 dB, lower is narrower and more resonant. |
attack_ms | Attack | 0.5 to 100 | 4 | ms | log | How fast each band envelope follows a rise in the modulator. |
release_ms | Release | 5 to 1000 | 60 | ms | log | How fast each band envelope falls after the modulator drops. |
sibilance | Sibilance | 0 to 1 | 0.2 | % | linear | Level of the modulator itself above High band, so consonants such as s and t stay audible. Stored as a fraction; shown ×100 as a percentage. |
formant | Formant | -12 to 12 | 0 | st | linear | Shifts the carrier bands up or down against the modulator bands, in semitones, for a smaller or larger voice. |
mix | Mix | 0 to 1 | 1 | % | linear | Blend from the dry carrier to the vocoded signal. Stored as a fraction; shown ×100 as a percentage. |
output | Output | -24 to 24 | 0 | dB | linear | Output trim after the mix. |
Bands (bands)#
8, 12, 16, 24 or 32 bands. Few bands sound vintage and blurry; many bands make words clearer. Changing the count crossfades over 5 ms.
Low band and High band (low_hz, high_hz)#
The centres of the lowest and highest bands. The bank always spans at least an octave and stays under 0.45 times the sample rate. Most of a voice sits between 100 Hz and 8 kHz.
Band width (width)#
How wide each band is relative to the spacing between bands. At 1, neighbours cross at −3 dB. Lower values make narrow, resonant, metallic bands; higher values overlap them for a smoother sound.
Attack and Release (attack_ms, release_ms)#
How fast each band follows the modulator when it gets louder and when it gets quieter. Short attacks keep consonants crisp; longer releases smear words into pads.
Sibilance (sibilance)#
Mixes in the modulator itself above the top band, through a 12 dB/oct high-pass. Consonants like s, t and f have little to vocode in most carriers; this keeps them audible.
Formant (formant)#
Moves the carrier bands up or down against the modulator bands, from −12 to +12 semitones. Down sounds like a larger voice, up like a smaller one. The pitch of the carrier does not change.
Mix and Output (mix, output)#
Mix blends from the dry carrier to the vocoded signal. Output trims the level after the mix.
Tips#
- A bright, harmonically rich carrier works best: a saw, a pulse, a chord or noise. A sine has nothing to vocode.
- Keep the modulator dry and up front. A voice with reverb on it smears every band.
- Raise Sibilance when words lose their edges, and lower it if the hiss is distracting.
- For a whisper, put noise on Audio.
See also#
Oscillator, Noise, Filter, Parametric EQ.
Help patches
Each help patch is an ordinary Toy that teaches one idea. The diagram shows the Patch; the preview shows its Frame. Open it in TuneToys to play and edit it.
1. Make a synth talk
beginner Teaches: bands, attack_ms, release_ms, mix
A saw wave vocoded by a noise voice whose formant wanders and whose syllables pulse.
MIDI1 lane
Audio2ch
Fundamental pitch, capped below Nyquist; detune uses cents. Range: 20–20000 Hz. Default: 220 Hz. Logarithmic scale. Units: Hz.
Hz
Output gain. Range: 0–0.5 ×. Default: 0.12 ×. Linear scale. Units: linear gain.
×
Continuous sine/triangle/saw/pulse morph; only its two neighboring waveforms are evaluated. Range: 0–3. Default: 0. Linear scale. Units: unitless. Marks: 0 Sine, 1 Triangle, 2 Saw, 3 Pulse.
Saw
Pulse duty cycle with a frequency-dependent safe edge width. Range: 0.05–0.95. Default: 0.5. Linear scale. Units: unitless.
Selects the implemented algorithm. Choices are discrete and are not LFO targets. Range: 0–7. Default: 0. Linear scale. Units: unitless.
Cycle offset; Gate rising edges restart phase from this position. Range: 0–1 cycles. Default: 0 cycles. Linear scale. Units: cycles.
cycles
Pitch offset in cents. Range: -1200–1200 cents. Default: 0 cents. Linear scale. Units: cents.
cents
On or off. A rising edge retriggers; held on runs the oscillator or envelope. A control cable switches it at 0.5. Range: 0–1. Default: 1. Linear scale. Units: unitless.
Semitones a full pitch bend on the MIDI input moves the note, up or down. Range: 0–48 st. Default: 2 st. Linear scale. Units: st.
st
Time constant of the slide from one MIDI note's pitch to the next; 0 jumps. Range: 0–2000 ms. Default: 0 ms. Linear scale. Units: ms.
ms
Audio1ch
Peak amplitude of white noise. Each channel and node path owns a repeatable independent stream; preparation restarts it. Range: 0–0.5 ×. Default: 0.12 ×. Linear scale. Units: linear gain.
×
Audio1ch
Audio1ch
Filter cutoff, capped at 0.45 times sample rate. Range: 20–20000 Hz. Default: 1200 Hz. Logarithmic scale. Units: Hz.
Hz
State-variable filter Q. Range: 0.5–12 Q. Default: 0.707 Q. Linear scale. Units: Q.
Q
Selects the implemented algorithm. Choices are discrete and are not LFO targets. Range: 0–5. Default: 1. Linear scale. Units: unitless.
One or two cascaded state-variable stages; notch/all-pass retain their response semantics. Range: 0–1. Default: 0. Linear scale. Units: unitless.
Reset1 lane
Control1 lane
Cycle1 lane
Waveform of the oscillator. The random shapes draw a new value each cycle from a generator seeded by the node path, so every host and replay agrees. Range: 0–5. Default: 0. Linear scale. Units: unitless.
Cycles per second when Beats is 0. Range: 0.01–40 Hz. Default: 1 Hz. Logarithmic scale. Units: Hz.
Hz
Beats per cycle, synced to the shared transport. 0 uses Rate instead. Range: 0–64 beats. Default: 0 beats. Linear scale. Units: beats.
beats
Peak-to-peak amount of the output swing. Range: 0–1. Default: 1. Linear scale. Units: unitless.
Centre of the swing. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Starting phase of the cycle, and where a Reset edge restarts it. Range: 0–1. Default: 0. Linear scale. Units: unitless.
Where the Triangle peaks, or how the Saw bends. 0.5 is the plain shape. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Reset1 lane
Control1 lane
Cycle1 lane
Waveform of the oscillator. The random shapes draw a new value each cycle from a generator seeded by the node path, so every host and replay agrees. Range: 0–5. Default: 0. Linear scale. Units: unitless.
Cycles per second when Beats is 0. Range: 0.01–40 Hz. Default: 1 Hz. Logarithmic scale. Units: Hz.
Hz
Beats per cycle, synced to the shared transport. 0 uses Rate instead. Range: 0–64 beats. Default: 0 beats. Linear scale. Units: beats.
beats
Peak-to-peak amount of the output swing. Range: 0–1. Default: 1. Linear scale. Units: unitless.
Centre of the swing. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Starting phase of the cycle, and where a Reset edge restarts it. Range: 0–1. Default: 0. Linear scale. Units: unitless.
Where the Triangle peaks, or how the Saw bends. 0.5 is the plain shape. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Audio1ch
Audio1ch
Multiplies every audio channel. 0 silences the signal; 1 leaves its level unchanged. Range: 0–2 ×. Default: 1 ×. Linear scale. Units: linear gain.
×
Audio2ch
Key1ch
Audio2ch
Number of bands, spaced evenly in pitch from Low band to High band. More bands make speech clearer. Range: 0–4. Default: 2. Linear scale. Units: unitless.
Centre of the lowest band. Range: 20–2000 Hz. Default: 100 Hz. Logarithmic scale. Units: Hz.
Hz
Centre of the highest band, kept at least an octave above Low band and below 0.45 times the sample rate. Range: 500–16000 Hz. Default: 8000 Hz. Logarithmic scale. Units: Hz.
Hz
Band width relative to the band spacing: 1 makes neighbouring bands cross at −3 dB, lower is narrower and more resonant. Range: 0.5–2 ×. Default: 1 ×. Logarithmic scale. Units: ×.
×
How fast each band envelope follows a rise in the modulator. Range: 0.5–100 ms. Default: 4 ms. Logarithmic scale. Units: ms.
ms
How fast each band envelope falls after the modulator drops. Range: 5–1000 ms. Default: 60 ms. Logarithmic scale. Units: ms.
ms
Level of the modulator itself above High band, so consonants such as s and t stay audible. Range: 0–100 %. Default: 20 %. Linear scale. Units: percent.
%
Shifts the carrier bands up or down against the modulator bands, in semitones, for a smaller or larger voice. Range: -12–12 st. Default: 0 st. Linear scale. Units: st.
st
Blend from the dry carrier to the vocoded signal. Range: 0–100 %. Default: 100 %. Linear scale. Units: percent.
%
Output trim after the mix. Range: -24–24 dB. Default: 0 dB. Linear scale. Units: dB.
dB
Audio2ch
Connections as text
- Oscillator audio output connected to Vocoder audio input
- Noise audio output connected to Filter audio input
- Control LFO “wander” control output connected to Filter @cutoff input
- Filter audio output connected to Gain audio input
- Control LFO “syllables” control output connected to Gain @level input
- Gain audio output connected to Vocoder key input
- Vocoder audio output connected to Patch outputs audio input
- The saw on Audio is the carrier: it supplies the pitch. The noise on Key is the modulator: its moving band-pass and pulsing level stand in for a voice.
- Each band of the modulator sets the level of the same band of the carrier, so the saw takes on the voice's moving vowel and rhythm.
- More bands make the words clearer. Attack and Release set how tightly the bands follow the voice; Mix 0 plays the plain saw.
2. Shift the formant, keep the consonants
intermediate Teaches: formant, sibilance, width, low_hz, high_hz, output
Formant moves the carrier bands, Sibilance lets the modulator's hiss through, and the band range and width shape the bank.
MIDI1 lane
Audio2ch
Fundamental pitch, capped below Nyquist; detune uses cents. Range: 20–20000 Hz. Default: 220 Hz. Logarithmic scale. Units: Hz.
Hz
Output gain. Range: 0–0.5 ×. Default: 0.12 ×. Linear scale. Units: linear gain.
×
Continuous sine/triangle/saw/pulse morph; only its two neighboring waveforms are evaluated. Range: 0–3. Default: 0. Linear scale. Units: unitless. Marks: 0 Sine, 1 Triangle, 2 Saw, 3 Pulse.
Saw
Pulse duty cycle with a frequency-dependent safe edge width. Range: 0.05–0.95. Default: 0.5. Linear scale. Units: unitless.
Selects the implemented algorithm. Choices are discrete and are not LFO targets. Range: 0–7. Default: 0. Linear scale. Units: unitless.
Cycle offset; Gate rising edges restart phase from this position. Range: 0–1 cycles. Default: 0 cycles. Linear scale. Units: cycles.
cycles
Pitch offset in cents. Range: -1200–1200 cents. Default: 0 cents. Linear scale. Units: cents.
cents
On or off. A rising edge retriggers; held on runs the oscillator or envelope. A control cable switches it at 0.5. Range: 0–1. Default: 1. Linear scale. Units: unitless.
Semitones a full pitch bend on the MIDI input moves the note, up or down. Range: 0–48 st. Default: 2 st. Linear scale. Units: st.
st
Time constant of the slide from one MIDI note's pitch to the next; 0 jumps. Range: 0–2000 ms. Default: 0 ms. Linear scale. Units: ms.
ms
Audio1ch
Peak amplitude of white noise. Each channel and node path owns a repeatable independent stream; preparation restarts it. Range: 0–0.5 ×. Default: 0.12 ×. Linear scale. Units: linear gain.
×
Audio1ch
Audio1ch
Filter cutoff, capped at 0.45 times sample rate. Range: 20–20000 Hz. Default: 1200 Hz. Logarithmic scale. Units: Hz.
Hz
State-variable filter Q. Range: 0.5–12 Q. Default: 0.707 Q. Linear scale. Units: Q.
Q
Selects the implemented algorithm. Choices are discrete and are not LFO targets. Range: 0–5. Default: 1. Linear scale. Units: unitless.
One or two cascaded state-variable stages; notch/all-pass retain their response semantics. Range: 0–1. Default: 0. Linear scale. Units: unitless.
Reset1 lane
Control1 lane
Cycle1 lane
Waveform of the oscillator. The random shapes draw a new value each cycle from a generator seeded by the node path, so every host and replay agrees. Range: 0–5. Default: 0. Linear scale. Units: unitless.
Cycles per second when Beats is 0. Range: 0.01–40 Hz. Default: 1 Hz. Logarithmic scale. Units: Hz.
Hz
Beats per cycle, synced to the shared transport. 0 uses Rate instead. Range: 0–64 beats. Default: 0 beats. Linear scale. Units: beats.
beats
Peak-to-peak amount of the output swing. Range: 0–1. Default: 1. Linear scale. Units: unitless.
Centre of the swing. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Starting phase of the cycle, and where a Reset edge restarts it. Range: 0–1. Default: 0. Linear scale. Units: unitless.
Where the Triangle peaks, or how the Saw bends. 0.5 is the plain shape. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Reset1 lane
Control1 lane
Cycle1 lane
Waveform of the oscillator. The random shapes draw a new value each cycle from a generator seeded by the node path, so every host and replay agrees. Range: 0–5. Default: 0. Linear scale. Units: unitless.
Cycles per second when Beats is 0. Range: 0.01–40 Hz. Default: 1 Hz. Logarithmic scale. Units: Hz.
Hz
Beats per cycle, synced to the shared transport. 0 uses Rate instead. Range: 0–64 beats. Default: 0 beats. Linear scale. Units: beats.
beats
Peak-to-peak amount of the output swing. Range: 0–1. Default: 1. Linear scale. Units: unitless.
Centre of the swing. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Starting phase of the cycle, and where a Reset edge restarts it. Range: 0–1. Default: 0. Linear scale. Units: unitless.
Where the Triangle peaks, or how the Saw bends. 0.5 is the plain shape. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Audio1ch
Audio1ch
Multiplies every audio channel. 0 silences the signal; 1 leaves its level unchanged. Range: 0–2 ×. Default: 1 ×. Linear scale. Units: linear gain.
×
Audio1ch
Peak amplitude of white noise. Each channel and node path owns a repeatable independent stream; preparation restarts it. Range: 0–0.5 ×. Default: 0.12 ×. Linear scale. Units: linear gain.
×
Reset1 lane
Control1 lane
Cycle1 lane
Waveform of the oscillator. The random shapes draw a new value each cycle from a generator seeded by the node path, so every host and replay agrees. Range: 0–5. Default: 0. Linear scale. Units: unitless.
Cycles per second when Beats is 0. Range: 0.01–40 Hz. Default: 1 Hz. Logarithmic scale. Units: Hz.
Hz
Beats per cycle, synced to the shared transport. 0 uses Rate instead. Range: 0–64 beats. Default: 0 beats. Linear scale. Units: beats.
beats
Peak-to-peak amount of the output swing. Range: 0–1. Default: 1. Linear scale. Units: unitless.
Centre of the swing. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Starting phase of the cycle, and where a Reset edge restarts it. Range: 0–1. Default: 0. Linear scale. Units: unitless.
Where the Triangle peaks, or how the Saw bends. 0.5 is the plain shape. Range: 0–1. Default: 0.5. Linear scale. Units: unitless.
Audio1ch
Audio1ch
Multiplies every audio channel. 0 silences the signal; 1 leaves its level unchanged. Range: 0–2 ×. Default: 1 ×. Linear scale. Units: linear gain.
×
Audio1ch
Mouth.in1ch
dB
Position on a two-lane mix: −1 left, 0 centre, 1 right. Constant-power for a one-lane channel, balance for a two-lane one; ignored on other widths. Range: -1–1. Default: 0. Linear scale. Units: unitless. The ring fills from zero for this signed linear range. Manual editing is unavailable while this host owns the control.
Read only
Sss.in1ch
dB
Position on a two-lane mix: −1 left, 0 centre, 1 right. Constant-power for a one-lane channel, balance for a two-lane one; ignored on other widths. Range: -1–1. Default: 0. Linear scale. Units: unitless. The ring fills from zero for this signed linear range. Manual editing is unavailable while this host owns the control.
Read only
Mix output level in decibels, after every channel. −60 dB is off; 0 dB leaves the sum unchanged. Range: -60–12 dB. Default: 0 dB. Linear scale. Units: dB.
dB
Audio2ch
Key1ch
Audio2ch
Number of bands, spaced evenly in pitch from Low band to High band. More bands make speech clearer. Range: 0–4. Default: 2. Linear scale. Units: unitless.
Centre of the lowest band. Range: 20–2000 Hz. Default: 100 Hz. Logarithmic scale. Units: Hz.
Hz
Centre of the highest band, kept at least an octave above Low band and below 0.45 times the sample rate. Range: 500–16000 Hz. Default: 6000 Hz. Logarithmic scale. Units: Hz.
Hz
Band width relative to the band spacing: 1 makes neighbouring bands cross at −3 dB, lower is narrower and more resonant. Range: 0.5–2 ×. Default: 1 ×. Logarithmic scale. Units: ×.
×
How fast each band envelope follows a rise in the modulator. Range: 0.5–100 ms. Default: 4 ms. Logarithmic scale. Units: ms.
ms
How fast each band envelope falls after the modulator drops. Range: 5–1000 ms. Default: 60 ms. Logarithmic scale. Units: ms.
ms
Level of the modulator itself above High band, so consonants such as s and t stay audible. Range: 0–100 %. Default: 50 %. Linear scale. Units: percent.
%
Shifts the carrier bands up or down against the modulator bands, in semitones, for a smaller or larger voice. Range: -12–12 st. Default: -4 st. Linear scale. Units: st.
st
Blend from the dry carrier to the vocoded signal. Range: 0–100 %. Default: 100 %. Linear scale. Units: percent.
%
Output trim after the mix. Range: -24–24 dB. Default: 0 dB. Linear scale. Units: dB.
dB
Audio2ch
Connections as text
- Oscillator audio output connected to Vocoder audio input
- Mixer audio output connected to Vocoder key input
- Vocoder audio output connected to Patch outputs audio input
- Noise “voice” audio output connected to Filter audio input
- Control LFO “wander” control output connected to Filter @cutoff input
- Filter audio output connected to Gain “mouth” audio input
- Control LFO “syllables” control output connected to Gain “mouth” @level input
- Gain “mouth” audio output connected to Mixer mouth.in input
- Noise “hiss” audio output connected to Gain “sss” audio input
- Control LFO “consonants” control output connected to Gain “sss” @level input
- Gain “sss” audio output connected to Mixer sss.in input
- A short burst of hiss joins the voice on every beat, like an s. Sibilance passes it above High band, where the saw has little to vocode.
- Formant shifts the carrier bands against the modulator bands: down for a larger voice, up for a smaller one.
- Low band and High band set the range of the bank; Band width overlaps the bands for a smoother, less resonant sound.
Related nodes
- Oscillator Selected waveform kernels, continuous morphing, harmonic shapes and phase/gate control; an optional MIDI input plays notes in the Toy's tuning.
- Noise Deterministic white noise with independent channel and instance streams.
- Filter Low-pass, high-pass, band-pass, notch, all-pass or morphing filter with resonance, 12/24 dB slopes, curve editing and output RTA.
- Parametric EQ Parametric EQ with up to 32 reorderable RBJ bands: cuts, shelves, bells and notches, per-band bypass and output trim.