Authenticate audio evidence.
Authenticate audio for investigations and disputes. A citable, reproducible verdict on whether a voice is AI-generated, backed by a versioned methodology.
Recordings enter evidence already suspect
Cloned-voice clips now surface in disputes, custody matters, employment cases, and fraud investigations, where a fabricated recording can shift a decision before anyone questions it. Investigators and counsel need a defensible way to ask whether a recording is synthetic, with a record of how the question was answered.
How it fits a caseA citable, reproducible analysis
Run the recording through the detector and save the verdict. You get a probability, a confidence level, the recognized model, a methodology version stamp, and a one-way audio fingerprint, all under a permanent citation URL. Because the methodology is published and versioned, an opposing expert can reproduce the test against the same snapshot. This is audio authentication in the practical sense: a documented, repeatable analysis of whether the voice was machine-generated. For the method, see how to verify AI audio.
Honest scopeAnalysis, not a ruling
We provide a probability with a published method behind it. We do not determine admissibility, we are not a law firm, and we do not certify identity; whether and how a verdict is used in a proceeding is for counsel and the court. We analyze recordings after the fact, delete audio within 24 hours unless a verdict is saved, and never claim a guaranteed 100% result. Integrate into a review workflow with the API, and for provenance see C2PA. The goal is narrow and defensible: a documented answer to whether the voice was machine-generated, produced the same way every time, that an expert on either side can examine.
Where confidence comes from
The verdict is a probability, never rounded up, paired with a confidence level driven by the length and quality of the recording. Clean, longer audio yields high confidence; short, compressed, or noisy clips yield less, and the detector reports that honestly rather than overstating certainty. For a contested recording, capture the clearest available copy and avoid re-recording it through another device, which adds compression and strips the artifacts the analysis relies on. Because the analysis is documented and repeatable, it supports a case file, a dispute, or a referral without asking anyone to trust a black box.
What every verdict gives you
Common questions
Is a verdict admissible in court?
Can the analysis be reproduced?
What do you store?
Does it identify whose voice it is?
Can we run it at volume?
Verify it before you rely on it.
Free for a single verdict. API for teams on Starter and above.