01
Disclose and ask
The call flow explains what will happen and asks for a clear yes. If the caller declines, enrolment does not run, the call continues the ordinary way and the decision is recorded.
Voice verification for Camlin Connect
Confirm a caller’s identity from their voice
Camlin VoiceID adds voice verification to the call flows you already run in Camlin Connect. A caller enrols once, after hearing what will happen and agreeing to it. On later calls VoiceID compares their voice to their own template while they speak, or after a spoken passphrase for higher-assurance steps. The call flow receives match, no match or inconclusive. It never receives a score.
- Consent captured before any enrolment
- One-to-one verification only, no watchlists
- Detects replayed recordings and synthetic voices
- Built by Camlin, hosted in Sydney or in your own environment
- Authored in Architect, runs inside Contact calls
VoiceID is in early access with pilot customers. Tell us about your call flow and we will say where it fits.
How it works
Three steps in a call
VoiceID runs as three actions in a Camlin Contact call flow, on the audio path the call already uses. No new IVR is needed.
02
Enrol from the live call
A few seconds of natural speech becomes a protected voice template. Audio that is too short, too quiet, clipped, noisy or contains more than one speaker is refused. A weak enrolment never becomes a credential.
03
Verify on the next call
The returning caller is compared one-to-one against their own template. The call flow receives match, no match or inconclusive and takes the path you authored: continue, step up, or hand to a person. Neither the caller nor the agent hears a score.
Where it fits
Where VoiceID is used
VoiceID runs at the steps you choose, such as an account change or a recovery call. It does not need to run on every call.
Account changes and recovery calls
Banks, telcos and insurers step up verification before high-risk actions instead of asking another round of security questions.
Government service lines
Confirm identity without asking people to repeat personal details aloud in public.
Agent-assisted verification
The agent desktop shows the result and the reason code. The biometric data stays in the engine.
Return callers and callbacks
Use the calling number as a hint only. Confirm with the caller’s voice and greet by name only after a match.
Voice plus screen
Send a Visual IVR link by SMS to capture name and consent on screen while the call continues.
What's in it
What VoiceID includes
The controls a security, fraud or contact-centre team will ask about.
Passive and passphrase modes
Verify from natural speech during the conversation, or ask for a randomised spoken passphrase for higher-assurance steps.
Consent evidence on every enrolment
The disclosure id, consent basis, capture path and claimed-identity source are recorded with the template.
Protected, revocable templates
Templates are keyed per store and can be reset or revoked for any individual. The stored credential is a template derived from the voice; it is not a recording.
Quality checks at enrolment
Enrolment fails closed on audio that is too short, too quiet, clipped, noisy or mixed-speaker.
Presentation-attack detection
Detects replayed recordings and synthetic voices on telephony audio. The result reaches the call flow as an outcome and a reason code, not a score.
Operations visibility
The operations console shows the enrolled roster, live verification events and voiceprint health. It shows results and reason codes; it does not expose biometric data.
Security & privacy
How VoiceID handles biometric data
Voice biometrics are sensitive information under the Privacy Act 1988 (Cth). VoiceID requires consent before enrolment and stores a template rather than a recording.
- Consent is captured before enrolment. The decline path is part of the call flow.
- The credential is a keyed template. It can be revoked for any person.
- One-to-one only. VoiceID does not search a caller against a list.
- Callers and agents receive outcomes. Nobody receives a score.
- Operational logs carry outcomes and reason codes. They do not carry audio or biometric vectors.
Biometric privacy notice
Recordings are used to build a voice template and are retained only as your consent and your deployment’s retention setting allow; templates can be revoked for any person at any time. How Nuamedia handles personal and sensitive information, including biometric templates, is set out in the Nuamedia privacy policy.
Hosting
Hosted in Sydney, or in your own environment
Camlin-hosted in AWS Sydney (ap-southeast-2), in Nuamedia’s own account, on an engine Camlin built. No third-party biometrics cloud is in the path.
Customer-hosted: run the same engine in your own environment, with the same consent, template and outcome rules.
Measurement
Accuracy is measured per deployment
Genuine and impostor attempts are run against enrolled voices under a published protocol and reported as ROC/DET curves and the equal-error point, following ISO/IEC 19795-1:2021, with drift flags between runs. We show you the method and your own results. We do not publish a single accuracy figure.
Works with Camlin Connect
How VoiceID fits the platform
VoiceID is authored as actions in Architect, runs inside Contact calls on the live audio, uses Speech for passphrase steps and reports into Insight.
- Camlin Architectauthors the actions
- Camlin Contactruns the call
- Camlin VoiceIDconfirms who is speaking
- Camlin Speechconfirms the words said
- Camlin Insightrecords outcome and reason
Camlin Architect
VoiceID is a set of call-flow actions: consent, enrol, verify, passphrase challenge, look up and reset. They can be placed in any flow. The engine credential never enters the flow definition or the browser.
Camlin Contact
The check runs on the live audio path the call already uses. No second recording leg is needed.
Camlin Speech
For a passphrase step, Camlin Speech confirms the words that were said. VoiceID confirms who said them.
Visit Camlin Speech →Camlin Insight
Every enrolment and verification is recorded with its outcome and reason code.
Visit Camlin Insight →
FAQ
Common questions
- Is my voice recording stored?
- The credential is a protected template, not a recording. How long any audio is kept depends on the consent given and the deployment’s retention setting. Templates can be revoked for any person.
- What if someone plays a recording of me?
- VoiceID checks telephony audio for replayed recordings and synthetic voices and reports a detected attack to the call flow with a reason code. We still recommend using VoiceID as one signal in a layered decision, alongside your existing controls.
- Can it identify an unknown caller?
- No. VoiceID is one-to-one only. It confirms a claimed identity against that person’s own template and does not search a caller against a list.
- What happens when it cannot decide?
- The call flow receives “inconclusive” and takes the path you authored: ask again, step up with a passphrase or knowledge questions, or hand to a person.
- Is it available today?
- VoiceID is in early access with pilot customers, hosted in Sydney. Tell us about your call flow and we will say where it fits.
- Can we run it in our own environment?
- Yes. The same engine can run in your own environment. The Camlin-hosted service runs in AWS Sydney, in Nuamedia’s own account.
- How accurate is it?
- Accuracy is measured per deployment with genuine and impostor testing that follows ISO/IEC 19795-1. We publish the method and you see your own results. We do not publish a single accuracy figure.
- Does it need a new IVR?
- No. VoiceID runs inside Camlin Contact call flows authored in Architect, on the live audio path the call already uses.
Talk to us
Book a briefing
Tell us one call flow where identity matters, such as a recovery call or an account change. We will walk through where VoiceID would sit in it, what the caller hears and what your team sees.
Already on Camlin Connect? Sign in and ask your account contact about the VoiceID actions in Architect.