Vocoder: Give a Carrier Rhythm and Articulation
The modulator supplies changing spectral envelopes; the carrier supplies the sound those envelopes shape. Keep those roles separate and Vocoder becomes much easier to control.
Follow the two roles through one device
Incoming audio is the modulator. Vocoder sums it to mono for analysis, divides it into frequency bands, and follows the amplitude in each band. Those envelopes shape matching bands of a carrier. You are transferring articulation and spectral emphasis, not simply shifting the modulator's notes.
The example supplies an original eight-second vowel-like synthesized phrase as the modulator. It contains no human voice, words or recorded singer. Saw selects the internal tonal carrier at 110 Hz, so this lesson needs neither a second track nor held MIDI notes. Play it fully wet and follow the repeated openings and closings of the tone.
Let a synthetic vowel envelope articulate a saw
Interactive audio example
Compare Carrier at 110 and 220 Hz. The carrier register changes while the timing still comes from the source. Then choose Noise: the envelopes now shape broadband noise instead of a fixed-pitch saw. Carrier is dimmed because its frequency applies only to Saw. Modulator uses the input itself as the carrier; it does not mean an external track selector.
Repair the envelope before adding detail
A blurred opening is an Attack question. Increase Attack from 5 ms toward 100 ms and listen for the start of each synthetic syllable becoming slower. That can turn a rhythmic sound into a pad, but it can also erase the small event that made the phrase intelligible as a rhythm. Return toward a faster attack when the beginning matters.
An overlong ending is a Release question. Compare the starting 80 ms with 400 ms. Longer release allows the carrier bands to persist as the modulator falls, filling the gaps. Shorter release makes the articulation more clipped. Listen at the boundary between events and stop lengthening it when consecutive gestures become one sustained wash.
These controls follow individual band envelopes, not a MIDI amplifier envelope triggered by notes. A quiet input can produce weak modulation, and the carrier alone is not a guarantee of audible output. If the result is faint, confirm that the source is playing before compensating with Gain. Avoid interpreting a large level boost as improved articulation.
Decide how detailed the spectral imprint should be
Bands spans 4–40. Compare 8 with 16 while keeping Range Lo at 80 Hz and Range Hi at 8 kHz. Fewer bands give a coarser spectral outline; more divide the same range more finely. They are not an automatic intelligibility setting. This synthetic material cannot demonstrate consonant recognition or prove how understandable a real spoken phrase will be.
Range Lo and Range Hi choose the analysis and synthesis region. Raising the lower edge removes low-band participation; lowering the upper edge excludes brighter detail. Bands and both range controls rebuild the filter bank, so make deliberate changes and compare a fresh phrase rather than treating their resets as musical envelope behavior.
Change the mouth-like color without changing the carrier note
Formant shifts the carrier band centers in semitones. Try +4 and -4 with Carrier unchanged; listen to the spectral color rather than expecting another melodic transposition. Res sharpens the bands. Too much can replace a broad articulation with narrow, insistent resonances, so lower it if individual bands begin to dominate the phrase.
Unvoiced blends noise into a tonal carrier and can add breathy high-frequency texture. It cannot invent missing consonants or lyrics in this source. Add a small amount, compare with zero, then use Dry/Wet if some original articulation should remain. This implementation has Noise, Saw and self-Modulator carriers only: external-carrier routing and MIDI-played carrier chords are not part of this panel or this demonstration.