←

Voice Biometric Rights

πŸ”’ All voice processing is 100% local. Data never leaves your device.

πŸ’‘ How to use Voice Biometric β–Ύ
Register and analyze voice biometrics for vocal-rights protection β€” 100% in your browser, nothing is uploaded. 1) Read the Biometric Consent Notice and tick the box to enable capture. 2) Upload an audio sample (WAV/FLAC/MP3) or record one with your microphone; the tool extracts a 192-d voice descriptor (ECAPA-TDNN) and runs an anti-spoof / liveness check (AASIST). 3) Compare a new sample against your registry to estimate similarity. 4) Optional: sign the analysis with your DID and timestamp it (OpenTimestamps) for provenance. 5) On the Manage tab you can export an encrypted backup, import it on another device, or delete entries. ⚠️ All processing is on-device. Withdraw consent at any time to delete every stored biometric record (GDPR Art. 7(3) + 17).

ECAPA-TDNN (192-dim) downloads an ~83 MB ONNX model on first use. WavLM Base+ (512-dim) downloads a ~102 MB model β€” best accuracy, heavier; embeddings stay local on this device.

Duration: 2–10 seconds of continuous speech gives the most reliable embedding; silent or music-only clips are rejected by the quality gate and flagged as low-confidence when processed.

⚑ Pick an audio sample or record one with the microphone, enter a Name/Label, then press Generate Identifiers β€” voice encoding, anti-spoof/liveness check, DID signing, registry (dis)similarity and provenance timestamp all run together automatically.

Registered voices: 0

Mixed embedding versions detected - re-embed older voice samples to enable cross-matching.

Deletes the consent record and all stored biometric data (GDPR Art. 7(3) + 17).

βš–οΈ Forensic similarity analysis β€” not a NIST-recognized voice authenticator. NIST SP 800-63B Β§3.2.3.2 prohibits biometric comparison based on voice for authentication; this tool reports similarity for evidence purposes only and never asserts an identity claim.

πŸ€– RedoSan Assistant