Vocal Resynthesis

The whole source-filter loop, live: SPEAK or SING into the mic and a synthetic Klatt voice reconstructs what you say — pitch, vowels, AND consonants. A NOISE GATE cleans the mic first — room noise, hum, and breath between phrases drop away below its Threshold (raise it if background still leaks, lower it if soft singing cuts) — so the analysers and the record only ever see your voice. Then three analysers hear you and drive one synth: the Pitch node tracks your f0 (drives the oscillator), the Formants node hears your vowel (drives the Voice's F1/F2/F3 filter), and the Manner node hears your excitation (drives the Voice's Voice/Breath/Frication mix) — an /s/ becomes hiss, a vowel voiced tone, a /m/ a murmur. An Env Follower on the gated mic opens the output VCA by your LEVEL — voiced OR unvoiced — so the synth follows your dynamics and a phrase that STARTS with a consonant (an 's…') isn't clipped, and it falls silent between phrases with no idle drone (the Follower's Attack/Release shape how snappy; the Manner Sensitivity dial trims breath/frication). Watch the Formants' vowel-plane and the Manner triangle light up as you talk. The reunification of the excitation arc: analysis -> resynthesis. Use headphones to avoid feedback (the mic's Mute also kills it). Grant mic access and say "sah-shoo-mee". A FUNCTIONAL preset (explicit nodes + connections), palette-only.

Signal chain