Skip to content
PRODUCTS / FACE MATCHING

Prove the holder is the person.

Biometric 1:1 comparison between a live selfie and the portrait on an identity document. One API call returns a calibrated similarity score, a pass / refer decision and explainable reason codes — in under a second.

1:1 biometric matchCalibrated 0–1 scoreTunable FAR / FRR99.7% accuracy
Emirates NBDMashreqe&duADCBCareem
99.7%
Match accuracy
<1s
Median decision
FAR / FRR
Tunable per use case
NIST FRVT
Benchmarked engine
Capabilities

One call. Every check.

Each capability runs in parallel and returns in a single structured response.

SIMILARITY

Calibrated scoring

A single 0–1 similarity score, calibrated so the same value means the same confidence across every camera and document — not a raw distance you have to interpret.

THRESHOLDS

Tunable FAR / FRR

Set your own decision threshold per flow, or pick a target false-accept rate and we map it to the score — strict for account opening, relaxed for re-authentication.

TOLERANCE

Pose & lighting robust

Matches reliably across off-angle faces, low light, glare, glasses and headwear, so real customers are not rejected for everyday capture conditions.

ANTI-FRAUD

Twin & mask resistant

Deep facial geometry separates identical twins and look-alikes, and flags silicone masks and prosthetics that defeat shallow matchers.

AGE DRIFT

Age-progression handling

Compensates for the gap between a document portrait and a current selfie, so a five-year-old passport photo still matches its holder.

EXPLAINABLE

Audit reason codes

Every result carries human-readable reason codes — score band, quality flags, attack signals — for your audit trail and your regulator.

Accuracy & thresholds

Decide where to draw the line.

The matcher returns a calibrated score, not a black-box verdict. You choose the operating point — tighten for high-value onboarding, loosen for friction-sensitive re-auth — and we hold the false-accept rate to your target as conditions change.

WHAT WE CHECK
  • Calibrated 0–1 similarity score, comparable across devices
  • Set a fixed threshold or target a false-accept rate (e.g. 1 in 100,000)
  • Per-flow profiles: strict onboarding vs. low-friction re-authentication
  • Quality gating returns a retry hint instead of a false reject
Anti-fraud

Built to defeat morphs and injection.

Presentation and injection attacks are scored alongside the match itself. Morphed portraits, photo substitution and virtual-camera feeds are flagged before they can pass as a genuine holder.

SAMPLE TEMPLATES
Morph detectionPhoto-substitutionMask & prostheticIdentical-twin separationVirtual-camera / injectionReplay detection
Developer experience

One call, structured result.

Send a selfie and the document portrait — or reference an existing verification session — and receive a typed response with score, decision and reason codes. SDKs for Web, iOS, Android, React Native and Flutter handle guided capture.

othento.js
const r = await othento.face.match({
  selfie: selfieImage,
  documentPortrait: idPortrait,
  threshold: 'onboarding',
});

// r.match            -> true
// r.similarityScore  -> 0.962
// r.decision         -> "pass"
// r.reasons          -> ["score_high", "quality_ok"]
How it works

Four steps, about a second.

01

Capture selfie

Guided SDK capture with live quality, framing and glare feedback — or pass an existing image.

02

Extract portrait

The reference face is cropped from the verified document, or reused from the active session.

03

Compare

The biometric engine compares facial geometry and scores presentation and injection attacks.

04

Decision

A calibrated score, pass / refer decision and reason codes are returned and logged.

Use cases

Onboarding, everywhere.

Banking

Account opening

Confirm the applicant matches their ID before a bank account or card is issued.

99.7% accuracy
Fintech

Wallet onboarding

Mobile-first selfie match that binds a real person to every new wallet without adding friction.

<1s to verify
Telecom

SIM registration

Match the subscriber to their national ID at activation, in-store or in-app, at scale.

60% faster
Why Othento

Manual review can't keep up.

Othento Manual review Legacy vendor
Time to decision Under a second Minutes per case Seconds–minutes
Score calibration Calibrated, comparable Subjective judgement Raw distance
Threshold control Per-flow, FAR-targeted N/A Single global cut-off
Twin & mask resistance Deep geometry Often missed Partial
Morph / injection signals Scored & flagged Not detectable Limited
Explainable reason codes On every decision Manual notes Rarely
SecuritySOC 2 Type IIISO 27001GDPReIDAS 2.0UAE PASS readyDIFC / ADGM residency
Othento Face Matching let us drop the manual selfie review queue entirely. We tuned the threshold to our risk appetite and the false-accept rate has held exactly where we set it.
Omar Haddad
Head of Fraud · Digital Bank
99.7%
match accuracy
-85%
manual reviews
FAQ

Questions, answered.

It is a calibrated value between 0 and 1, where the same number reflects the same confidence regardless of the camera or document used. You decide the threshold at which a score becomes a pass, or give us a target false-accept rate and we map it for you.

Get started

See Face Matching on your documents.