About
We work on one thing: the voice on a phone call.
A cloned voice is cheap to make and a phone line is where organisations still act on what a voice says. Trueline tells your team, about twelve seconds into a call, when the voice on the line was made by a machine.
- What we make
- Detection of cloned voices and AI voice agents on phone calls
- Who it is for
- Telecom, banking, insurance and contact centres
- How it connects
- A copy of the call audio, never in the call path
- What we keep
- Scores and verdicts. Never audio.
Focus
Not Video meetings.Not Social clips.Not Voice messages.Not Broadcast audio.Phone calls.
A synthetic voice can reach an organisation five ways. A phone network changes audio in ways the others do not, so a detector built for clean files does poorly on a call. We built ours for the call, and we do not claim the rest.
The call routes we measured →How we work
Four rules, each one checkable on this site.
- 01Measured on the route, not on the bench
- Every figure we publish comes from calls placed over real carrier and internet routes, through the same pipeline a customer's calls go through.
- Detection performance →
- 02The routes we are weak on are published
- Where a call path defeats the detector, the accuracy page says so, with the sample size beside it. You should hear it from us before a pilot, not after.
- What we do not claim →
- 03Never in the way of a call
- Trueline reads a copy of the audio. It adds no latency, it cannot drop a call, and when it is unavailable your calls carry on.
- How it works →
- 04Nothing kept that we do not need
- Audio is held in memory for at most eight seconds and never written to disk. There is no transcript and no voiceprint of your customers.
- Security →
The quickest way to judge us is a call.
Thirty minutes: we place a real call, play a synthetic voice down the line, and you watch the flag arrive. Bring your security team's questions.