WAVEGUIDE VOICE ENGINENatural voice from pure physics.
Natural voice from pure physics.
Zero dependencies.
No neural networks. No recordings. No training data. No runtime dependencies. Pulse Voice is a digital waveguide that produces human-quality speech from mathematics alone. Shape sound by manipulating the geometry of the vocal tract — formants emerge from physics, not lookup tables.
22050 Hz
sample rate
10
tract sections
0
heuristics
LF
glottal model
How It Works
LF Glottal Source
Liljencrants-Fant model generates airflow pulses. Rd controls voice quality from pressed to breathy. Jitter and shimmer add natural perturbation.
Waveguide Reactor
10-section digital waveguide with Kelly-Lochbaum scattering junctions. Cross-sectional areas define the vocal tract. Formants emerge from the geometry.
Nasal Coupling
Velum port connects a nasal side-branch. Opening it introduces nasal resonance (~250Hz) and anti-formants. Physical nasality from anatomy.
Phoneme Geometry
Every phoneme has a measured area function from MRI data. Load a phoneme preset and the tract reshapes to produce that sound. Pure physics.
Discover & Share
Found an interesting sound? Click “Share Preset” in the board to get a URL that encodes your exact settings. Share it with anyone — they'll hear exactly what you heard. Contribute your discoveries to the community tuning effort.
Licensing Verticals
The entire voice engine fits in a 2 MB kernel with zero runtime dependencies. Each vertical is a standalone licensable product built on the same waveguide physics.
