Product
Sonavera is a quick check that runs in the browser and returns signed evidence a real, live person was there.
Every Sonavera check is the same short, guided browser session. What changes is the question you ask: Is this the same person who enrolled? Is a live person present? Is this a unique person for a specific event? You pick the flow; we return the evidence.
Sonavera is not a login provider, not document KYC, and never makes the final call on your users.
Pick the question first
Each flow answers one question and refuses to pretend it answers more. That restraint is the point: you always know exactly what the evidence says, and what it doesn’t.
The user records face and voice once, inside your tenant. That baseline is what later verifications compare against.
Boundary: An enrollment exists only if the check actually captured enough data, and it expires on the retention schedule you configure and disclose.
A fresh session measures how closely today’s face and voice match the enrolled baseline, alongside liveness and media signals.
Boundary: This is similarity evidence inside your tenant. It doesn’t establish legal identity, and it doesn’t decide what you should do.
A one-off session gathers liveness, challenge, and media evidence tied to a single action. Nothing is saved for future comparison.
Boundary: Liveness never creates a biometric record and never claims who the person is.
Within a scope you define (a promotion, a waitlist, a beta), measure how distinct this person is from earlier participants.
Boundary: Scoped to your tenant and your time window. Not a global search, and prior participants are never exposed in your results.
Evidence, not verdicts
It’s tempting to collapse a check into one green or red light. We don’t, because the interesting cases live in between. Every result keeps four facts apart:
The user either completed the guided check or didn’t. That fact stands on its own.
Cameras fail and microphones cut out. The result says plainly whether we captured what the flow needed.
For enroll: was a baseline actually saved? For verify: did the comparison run? Doing and measuring are different facts.
The Evidence Index and its per-check breakdown, measured under a fixed, versioned profile.
Evidence Index
The index is composed from fixed allocations that ship with the profile. If a measurement is missing, its points simply aren’t earned, never quietly shuffled somewhere else. Open any result and you can trace the whole number back to what was actually observed.
measurements_partialStraight answers
Every signal below feeds evidence into the same result. None of them, alone or together, turns into a verdict about a person.
What it measures:The user responds to live prompts during the session.
Where it stops:Evidence a responsive person was there, not proof of who they are.
What it measures:How closely this session’s face and voice match the enrolled baseline, inside your tenant.
Where it stops:Not document KYC. Says nothing about legal or government identity.
What it measures:Signals that the video and voice are a live capture rather than a replay or a synthetic feed.
Where it stops:Strong evidence, not a guarantee against every attack ever invented.
What it measures:Whether reliable samples of the selected primary face remained mutually consistent through the session.
Where it stops:Missing or ambiguous observations can make evidence incomplete but do not directly lower the score. Not identity, liveness, continuous presence, or a count of people in the room.
What it measures:Coarse observations about the connection and browser, where the flow includes them.
Where it stops:Context alongside the evidence; these never add points to the Evidence Index.
What it measures:Whether the session-specific spoken phrase was observed.
Where it stops:The result keeps the fact, never the audio, never a transcript.
Product questions
No, and that’s deliberate. We compare a live person with a baseline your application enrolled, and report similarity evidence. Whether that account belongs to a particular legal identity is a fact your system owns. Document-based KYC checks documents. We check for real, live human presence.
No. Completion just means the session finished. The evidence inside can be strong, weak, partial, or missing, and the result says which, in plain fields. That honesty is what makes it safe to build policy on.
A 0–100 measure of how much supporting evidence the check gathered, composed under a fixed, versioned profile. It isn’t a probability, it isn’t a percentage, and it never comes with a Sonavera pass/fail label. Missing measurements keep their allocation and earn zero.
As long as your configured retention window allows. You choose the window, the user sees it before consenting, and deletion is scheduled from the start. A verification needs an active enrollment inside that window.
No. It compares only within a scope you create (one promotion, one waitlist) inside your own tenant and its bounded window. There is no cross-customer pool to search.
Add a live-person step without replacing your identity provider, your risk engine, or your judgment.