[ 01 ]
What makes cloned speech detectable
Generated speech is built differently from air-recorded speech. It can lack the room and microphone behaviour of a real capture, or show unusually smooth transitions. Splicing real and synthetic audio leaves joins in noise level and frequency content.
[ 02 ]
Core measurements
Spectral continuity checks whether frequency content evolves naturally. Noise-floor continuity checks that background sound is consistent across the recording. Network hum analysis looks at mains-frequency traces where available. Re-encoding analysis looks for signs the audio was compressed more than once.
[ 03 ]
Limits to state clearly
Telephone codecs and messaging apps remove detail, and heavy compression can hide or mimic these signs. A responsible report records the channel assumptions and its confidence.
[ 04 ]
Where it is used
Call-centre fraud, executive impersonation, leaked recordings in news verification and evidence in legal matters.
Quick answers
Can a cloned voice be detected from a phone call?
Often it can be examined, but phone codecs reduce detail, so the report states the limits of what the recording allows.
Does SpearTrace analyse live calls?
The platform analyses recordings and files. Results can be produced after the call.