Make a kubyz speak your words

A kubyz (jaw harp) already sounds like a mouth: you shape its tone with your lips and tongue. This recipe hands that job to your voice. You sing, and the kubyz pronounces your vowels. None of your voice is heard; every vowel is the kubyz.

What you need

• SuperDuper Formant, one of the free SuperDuper DSP plugins (macOS and Windows).
• A kubyz recording or the SuperDuper Kubyz instrument, and a vocal take of the same length.
• Any DAW with sidechain routing. I used REAPER.

The chain

1. Kubyz on a track. On it, Formant, then a reverb.
2. Send the voice to Formant’s sidechain input (it is called Voice). Mute the voice track itself.
3. Formant settings:

4. Reverb (I used Supermass) at Mix 0.22.

When the voice stops, the last vowel freezes and the kubyz keeps “holding” it. That is a feature: sing a phrase, stop, and the instrument finishes the sentence.

What to ask the assistant

Put the kubyz on one track and my vocal on another. Make the kubyz pronounce the vowels of the vocal with SuperDuper Formant in Follow mode, and mute the vocal.

To render it without a DAW, the same chain as a file for sdsp-chain:

sidechain = "voice.wav"

[[track]]
name = "kubyz articulated by voice"
input = "kubyz.wav"

  [[track.stage]]
  plugin = "formant"
  params = { Mode = 1.0, Follow = 1.0, Glide = 22.0, Width = 0.9, Drive = 0.3, Mix = 0.95, Output = 6.0 }

  [[track.stage]]
  plugin = "supermass"
  params = { Mix = 0.22, Width = 1.0 }

Why it works

A vowel is three resonances of the mouth (formants). Formant measures them on your voice many times a second and puts the same three resonances on the kubyz. The kubyz gives the pitch and the buzz, your voice gives the shape. Drive goes first because a plucked drone is too pure: without extra harmonics there is nothing for the resonances to shape.

More: project-superduper-dsp, the kubyz as a phone app in project-kubizbeat, and the whole setup in creative-music-with-claude.