Verify whether a voice memo, audio file, or call recording is a real human or an AI voice clone synthesized by ElevenLabs, Bark, or deepfake speech models.
What to listen for
Modern neural TTS models mimic human tone and inflection, but leave measurable acoustic fingerprints in the frequency spectrum:
Neural voice generators often show sharp frequency cutoffs above 16 kHz or artificial harmonic flattening between 2 kHz and 8 kHz.
Human speakers take natural breaths and produce vocal fry. AI audio often speaks continuously without natural inhalations.
Synthetic voice clones frequently exhibit identical pitch contour inflections across separate sentences.
Sounds like /p/, /b/, /s/, and /th/ often lack realistic mechanical friction against teeth and lips, sounding overly crisp or slightly muffled.
Targeted forensic checklists, generator signature scanners, and catfish verification guides.
Free, private, and deleted after your scan.
