Melody / Help / Process / Klatt Voice
Klatt Voice
The source-filter VOICE — turn a bare oscillator into a talking/singing voice. Where the Formant Filter shapes a carrier into a VOWEL (the vocal-tract FILTER), this adds the excitation SOURCE the vowel filter can't make on its own: a VOICE amount (the glottal carrier), a BREATH amount (aspiration noise through the tract — /h/, breathiness), and a FRICATION amount (turbulent hiss added after the tract — /s/, /ʃ/, /f/), so the voice can make CONSONANTS, not just vowels. The reunification of the source-filter model: wire a saw oscillator (its pitch from a Keyboard or the Pitch node) into the input for f0, drive the vowel from a Formants node's output (formants_in), and drive the source mix from a Manner node (source_in) — sing into a mic and a synthetic voice reconstructs what you say, consonants and all. Or play it directly: dial Voice/Breath/Frication + F1/F2/F3 by hand. Built on the same parallel-bandpass formant bank as the Formant Filter; aspiration is noise shaped by the vowel tract, and frication is a PARALLEL 5-band noise-formant bank driven by the Manner node's measured frication fingerprint, so /s/ (high) vs /ʃ/ (mid) vs /f/ (diffuse) render as DISTINCT consonants rather than one generic hiss. A separate BURST branch (place-steerable bandpass'd noise, a fast-envelope gain) fires a crisp plosive transient — /p t k b d g/ — the continuous drive can't make. For VOICED fricatives (/z v ʒ/) the frication noise is amplitude-MODULATED by the glottal flow (the voiced source, rectified, drives the noise gain) so buzz + turbulence pulse together at f0 as one coherent sound — and the voicing amount auto-gates it, so a voiceless /s/ stays steady. The DSP shape is kernel-declared (a web_audio graph); audio.js realises it.
Inputs
| Port | Signal |
|---|---|
| audio_in | audio |
| Vowel | audio · formant |
| Manner | audio · source |
Outputs
| Port | Signal |
|---|---|
| audio_out | audio |
Parameters
| Parameter | Range | Default |
|---|---|---|
| Voice | 0 – 1.5 | 1.0 |
| Breath | 0 – 1.0 | 0 |
| Frication | 0 – 1.5 | 0 |
| F1 | 200 – 1200 Hz | 730 |
| F2 | 600 – 3000 Hz | 1090 |
| F3 | 1500 – 4000 Hz | 2440 |
| Tilt | 800 – 6000 Hz | 3500 |
Used in these rigs
Klatt Voice · Vocal Resynthesis · Voice Sculptor
More process nodes
3-Band EQ · 3D Orbit · 3D Spatial · Amp Model · Auto-Pan · Band-pass Filter · Bitcrush · Buffer Stutter · Cabinet · Chorus · Compressor · Delay