Glyph VoiceMark - VoiceMark

The Glyph VoiceMark is the persistent identity portrait for a speaker. Unlike the Cloud VoiceMark, it is deterministic — the same voice always produces the same glyph, regardless of what phrase is spoken.

What it represents

A stable biometric identity mark — think of it as a voice fingerprint rendered visually
Appropriate for profile pages, enrollment confirmations, identity badges
Consistent across sessions; changing only if the voice significantly changes (illness, ageing)

Architecture

The glyph is derived from a 192-dim ECAPA-TDNN speaker embedding via a fixed random projection. The projection maps embedding dimensions to 10 visual parameters, each in [0, 1]. The portrait renders as a 12×12 concentric ring structure. Each ring’s shape, density, and colour is controlled by a subset of the glyph parameters.

Where it comes from

glyph_params is returned in two places:

/v1/enroll response — for immediate rendering after enrollment
/v1/verify response inside cloud_data.glyph_params — confirm the glyph on each verify

Glyph parameters

Key	Range	Visual role
`hue`	0–1	Primary colour of the concentric ring portrait
`shapeStretch`	0–1	Eccentricity of the outer ring
`shellDensity`	0–1	Number of rings rendered
`breathPulse`	0–1	Amplitude of the ambient pulse animation
`fluxMotion`	0–1	Speed of ring rotation
`scatterAir`	0–1	Particle scatter density in the outer field
`chaosWarp`	0–1	Ring distortion / irregularity
`pulseDepth`	0–1	Depth of the pulse effect
`kurtosisIntensity`	0–1	Peak sharpness of colour transitions
`colorVibrancy`	0–1	Saturation of the accent colour

Example

{
  "glyph_params": {
    "hue": 0.61,
    "shapeStretch": 0.44,
    "shellDensity": 0.73,
    "breathPulse": 0.38,
    "fluxMotion": 0.55,
    "scatterAir": 0.21,
    "chaosWarp": 0.18,
    "pulseDepth": 0.47,
    "kurtosisIntensity": 0.62,
    "colorVibrancy": 0.80
  }
}

⌘I