REAL TALKBOX EFFECT · LPC VOCAL TRACT MODELING, MIDI-PLAYED CARRIER
TALKBOY
Sing into it. The synth does the talking.
Not a vocoder — a real talkbox, done the way the mechanism actually works. TALKBOY listens to your voice, estimates the all-pole filter your vocal tract is making right now via linear prediction, and pushes a MIDI-played carrier through that exact filter instead of through your vocal cords. Five carriers (SAW, STACK, SUPER, PULSE, NOISE) cover the classic talkbox voices, and an EXT mode skips MIDI entirely and takes a real instrument on the sidechain — an actual instrument-into-mouth talkbox, patched in.
- Linear-prediction vocal tract estimation, not a vocoder — no fixed band grid, poles land wherever your formants actually are and slide continuously as your mouth moves
- Five pitched carriers (SAW, STACK, SUPER, PULSE) plus a NOISE whisper mode — SUPER is a from-the-die-accurate JP-8000 SuperSaw, seven naive saws with Roland's own detune law
- EXT mode needs no MIDI at all — patch a Rhodes, guitar, or synth bus into the sidechain and play a genuine instrument-into-mouth talkbox, gated on signal presence
- The synthesis filter runs in reflection-coefficient space, not direct-form — coefficients update every 4ms and can't pass through an unstable filter no matter how fast the vowels move
- KEY GATE decides what happens between phrases: THRU passes your dry voice bit-for-bit at zero cost, GATE goes silent so TALKBOY is the whole sound
- BREATH and AIR hand your voice's own sibilants and top octave back untouched — nothing in the vocal-tract model can produce a fricative or content above 8kHz, so none of it is invented
- Bit-exact at MIX 0 and with no note held, on every carrier and every sample rate — verified against the literal input, not just measured close
VST3
AU
STANDALONE
WIN / MAC
NO PRESETS · SPEND TIME WITH IT
