The tell is in the line itself.
A live caller keeps the signal breathing with its own irregular rhythm. Pipe in a generated voice and the same line goes mechanically regular — too even to be breathing.
We don't ask the caller to do anything. We just listen for the things only a live person produces.
Breath & micro-pauses
The small, uneven pauses and breaths of someone actually talking.
Room & background
The live acoustics around a real caller, versus audio generated in a clean signal chain.
Micro-timing
The natural, non-repeating rhythm generated speech can't sustain.
Vocoder signature
The faint vocoder signature left behind when speech is generated.
The moments where "is this a real person, right now?" is the whole question.
Account takeover
A cloned or synthetic reply shouldn't pass as a live customer on the line.
Voice passphrases & IVR
Flag synthetic voices fed into automated voice prompts and self-service menus.
High-value transfers
Confirm a real person is authorizing — in the moment, not after the money moves.
No passphrase. No prompts. No friction.
The caller simply speaks. Liveness runs continuously, in real time, and pairs with voice biometrics — biometrics confirms who is speaking, liveness confirms that a live person is speaking right now.
Runs inside your walls.
Voice Liveness deploys on-premise. Call audio and analysis stay within your own environment and never leave it.