Voicd Research

Secure. Ethical. Trusted.

Voicd focuses on how speech is produced, not just how it sounds, enabling structured, interpretative analysis across use cases. 

Grounded in Research

  • Physiology-first modeling of speech production

  • Independent, high-dimensional feature systems

  • Focus on interpretability and explainability

  • Validation across diverse datasets and real-world conditions

Our Research Focus

Security

Voice In The Past

  • Detects surface-level audio artifacts (e.g., glitches, unnatural smoothness, training signatures) 

  • Signals degrade as generative models improve 

  • Limited reliability on high-quality, real-world samples 

Voicd in The Future

  • Models the biology of speech production 

  • Evaluates consistency with human physiology 

  • Identifies sources of mismatch 

Designed to detect physiological inconsistencies in synthetic speech that artifact-based methods miss

Validation

Assessed for accuracy, consistency, and known error rates according to Daubert standards. Accepted as scientific evidence in a U.S. court 

Healthcare

Voice In The Past

  • Voice signal established as an indicator of certain neurodegenerative diseases (e.g. Parkinsons or ALS) 

  • High accuracy within small study datasets, but limited generalizability to real-world populations 

Voicd in The Future

  • Models the physiological system that produces speech 

  • Emphasizes underlying physiological mechanisms rather than dataset-specific correlations 

    Voicd provides a mechanism-based foundation for models to generalize across populations and real-world conditions

Status

Health applications are currently in the clinical research and validation phase. Commercial use will be pursued after securing regulatory approval