AI Audio Detector
Drop in an audio file and we'll run it through a series of independent checks — signed Content Credentials, SynthID watermarks, embedded generator tags, then the shape of the audio itself.
When a file carries a marker proving it was generated, we say so plainly. When it does not — which is most of the time — detection is a matter of probability, and we show you every measurement behind the estimate rather than a single number to take on faith.
Free to use, no account needed — 2 checks a day. Sign up and that goes up to 25.
Drop an audio file here
WAV, MP3, FLAC, AIFF, OGG, M4A · up to 100 MB
What this can and cannot tell you
- A confirmed result is reliable. Content Credentials are cryptographically signed, and a SynthID watermark is embedded at generation. Neither appears by accident.
- A clear result is not proof. Watermarks and metadata are easy to strip, and the SynthID check only covers audio from OpenAI's models — not Suno, Udio, Stable Audio or ElevenLabs.
- Short clips are hard. Below two seconds there is not enough audio for the signal measurements to mean anything, so we tell you that instead of guessing.
- Lo-fi is not AI.Vinyl samples, tape and field recordings share some characteristics with generated audio. We weigh those signals against the file's format before drawing any conclusion.