Skip to content
Guardian Line
Live Monitor

See it catch a cloned voice

This runs the same real audio-forensics engine used in the Live Monitor — pitch jitter, amplitude shimmer, and spectral flatness — entirely in your browser. Record a few seconds of your own voice, then generate a synthetic reference sample engineered to exhibit the acoustic fingerprints of AI-generated speech (near-perfect pitch stability, a mechanically regular amplitude envelope). Watch the scores diverge.

Your voice

Record ~3.5 seconds of yourself speaking naturally.

Synthetic reference sample

Procedurally generated with near-zero pitch jitter and a flat spectral envelope — the acoustic signature our forensics module looks for.

Honest framing: this reference sample is a procedurally generated test signal engineered to exhibit synthetic-speech acoustic properties — not a clone of a specific person's voice. It exists to demonstrate the forensics engine's real signal processing (autocorrelation-based pitch tracking, an in-house FFT for spectral analysis) against genuinely analyzable audio, without depending on a third-party voice-cloning model.