Overview

The phoneme table for sound.speech: about 40 English phonemes in ARPAbet, with the acoustic targets the synthesizer renders. Values start from the published Klatt (1980) tables and classic male formants. They are tuning data: adjust them by ear with tests/sound/manual/speak.lua.

table holds the phonemes by name, SIL among them.

Fields per phoneme; a missing field takes its type's default:

  type         "vowel"|"diph"|"stop"|"affr"|"fric"|"nasal"|"liquid"|"glide"|"asp"|"sil"
  f            {F1,F2,F3} formant targets, Hz (consonants: loci the transitions aim at)
  fe           {F1,F2,F3} diphthong end targets (diphthongs only)
  bw           {B1,B2,B3} bandwidths, Hz
  dur          inherent duration, ms (stressed); unstressed = dur * 0.65
  av           voicing amplitude 0..1
  ah           aspiration amplitude 0..1
  af           frication amplitude 0..1
  fa           {a2,a3,a4,a5,a6,ab} parallel frication spectrum amps
               (a2..a3 ride the live F2/F3; a4..a6 fixed high resonators; ab = bypass)
  fnz          nasal-zero frequency, Hz (nasals; pole is fixed at 280)
  stop         {closure=ms, burst=ms, asp=ms} stop-gap timing (stops/affricates)
  burst        fa-style spectrum for the release burst (stops)
  release      the fricative an affricate releases into, by name (affricates)
  release_dur  the release's length, ms (affricates)
  rank         boundary dominance (higher rank fixes the formant value at a boundary)
  trans        formant transition time at this phoneme's boundaries, ms

Functions