Truthring
Check a recording
Product
AI voice detector Deepfake audio detection Voice clone detection Audio verification Developer API
Solutions
Individuals and families Journalists Researchers Enterprises Security teams Contact centres Legal teams
Research
Methodology Accuracy Benchmark Limitations Research
Learn
What is AI voice detection? How to tell if a voice is AI How voice cloning works AI voice scams Learning centre
Resources
Glossary Free tools Comparisons Blog Pricing About
Check a recording
Prototype · not yet accepting submissions

Check a clip

Ten to thirty seconds of one person speaking gives the best result. A short clean section beats a long messy file.

Before a clip can be submitted

These are terms, not a formality. Submitting a recording you have no right to submit, or using a result as sole grounds for a decision about a person, is outside acceptable use and ends access.

Before you upload anything. Only submit a recording you have the right to submit. In many places a voice recording is personal data, and in some it is biometric data — which means analysing someone else’s voice can carry obligations even when you obtained the file lawfully.

A result here is a probability produced by a stated method with a stated error rate. It is not proof, not an expert opinion, and not evidence of anything on its own. Do not make it the sole basis for accusing, disciplining, dismissing or prosecuting anyone. What we do and do not claim is set out in limitations, acceptable use and compliance.

This service is a prototype. It is not open for submissions and is not a substitute for legal, financial or security advice.

Drop a clip, or choose a file Audio will be deleted once analysis finishes, and a one-way hash retained so a report can be tied to the exact file — never the file itself. [VERIFY: retention window not yet set] MP3 · WAV · M4A · OGG · WEBM · up to 25 MB

Best input

Ten to thirty seconds of one person speaking, from the original file rather than a forwarded copy.

Lowers confidence

Music, crowd noise, multiple speakers, heavy compression, and audio edited after recording. We say so when they apply.

What we keep

Nothing but a one-way SHA-256 hash of the file. The audio is deleted when analysis completes. Privacy

Reviewed