Voice synthesis
Checks usable speech for patterns associated with generated voices.
Analyze audio for AI signals
Use this check for voice notes, recordings and audio clips. Results depend on the kind and quality of sound available, so the report can return not applicable instead of inventing certainty.
Specialist checks Data handling explained Uncertainty shown
Check this source What this detector reviews
Media classification
Separates speech, music and unsuitable samples where supported.
DeepfakePolicy assessment
Returns the available model evidence without exposing processor storage links or identifiers.
Reading the result
Use the score as a signal, not proof
Not applicable means the detector did not have the right kind of signal to judge - it is not the same as authentic. Preserve the original and compare it with verified recordings when identity matters.
Important limitation: Instrumental sound, tones, silence and distorted recordings may not contain enough speech evidence for a voice detector.
Read the full methodologyWhen this check can help
Voice notes
Review an unexpected request that uses a familiar voice.
Podcast clips
Triage a disputed excerpt before publication.
Audio research
Pair model evidence with an original recording and source history.