V-DNA (VoiceDNA) detects AI-generated voice clones in real time using 59 biological audio markers—MFCCs, spectral entropy, pitch stability, breathing patterns, and harmonic structure. Trained on 7,356 samples across 9 AI systems, it achieves 90.15% F1 with a 0.11% false negative rate, optimized for real-world phone audio via G.711 simulation.
SooryaDayal/V-DNA
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|