Will extra checks hurt conversion?

Document, liveness, face match and device signals run in one onboarding scenario, not several separate forms. Ambiguous results can trigger re-capture or manual review instead of automatically rejecting honest customers.

What if the device looks suspicious?

Device Intelligence returns session signals to your flow. VPN or proxy is context, not an automatic decline: you define the response based on other results.

Can you detect multi-accounts?

Yes. Face Search 1:N compares the new face against your connected database of verified users. Average search speed is about 1.5 seconds — that is search time, not full onboarding duration.

Is AML mandatory in every scenario?

No. AML screening connects when needed and per your product requirements. On this page it remains an optional layer, not the center of onboarding.

Can we skip full KYC again in another product?

Reusing results from a completed verification is a possible scenario, not a standalone product promise on this page. On a demo we will walk through what data and steps can be reused in your architecture.

What does the user see on a weak device or bad capture?

Your flow can request re-capture; ambiguous results can go to manual review. One bad frame should not automatically mean a final decline for an honest customer.

Where is data stored?

Cloud deployment and on-premise options are available. The chosen model is fixed in architecture and contract. On a demo we clarify access, retention and storage rules for your product.

Create a cabinet: 500 checks a month free, no card. Or request a demo — we will review your onboarding and anti-fraud rules.

Biometrics for fintech

Honest customers move fast. Face, video and device spoofing risks are visible before opening the product.

Onboarding must do two things at once: keep honest customers moving and flag suspicious sign-ups before they access money or services. Document, liveness and face match are complemented by video spoof analysis and device signals — all in one product flow.

Start free Request a demo

Session · fintech_onboardingSuccess

matchedface ↔ document

Document captureddata extracted

Liveness checkpassed

Video spoof analysisno signs found

Device signalsin verdict

Session context · runtime, network, model and OS

Product can openVerdict returns to your onboarding flow

Sound familiar?

Fintech grows fast. Fraud walks through the same door.

Real customers sign up alongside bots, drops, stolen documents, synthetic identities and deepfakes. Selfie-to-document match alone is no longer enough: video can be spoofed, the device may look suspicious, and the same face may return for new accounts.

Tighten checks blindly and conversion suffers. Loosen them and fraud gets product access. The goal is not another barrier — it is to pass honest customers quickly and stop risk in time.

Registration streamlive

Registration 4831Deepfake signs on camerarisk

Registration 4832Wrong document · droprisk

Registration 4833Suspicious session environmentreview

Registration 4834Document · liveness · face · deviceapprove

Honest customers and fraud share one stream — split them before product access.

How verification works

From document to verdict — one scenario without extra forms

Document, liveness, face match, spoof analysis and device signals run in one flow — without a separate anti-fraud questionnaire. Each layer answers its own question: who is on camera, are they live, is video spoofed, can the device be trusted.

Document

The customer photographs a passport or ID. The system recognizes the document and extracts data for verification.

Step 1 · documentCaptured

Document typerecognized

Document dataextracted

Portrait from documentready for match

Document in sessionData available for the next layer

Liveness and face match

The system confirms a live person on camera and that their face matches the document photo.

Step 2 · liveness and faceSuccess

Selfie

Document

Live person on cameraconfirmed

Face match with documentmatched

Face confirmedMatch threshold configurable to your risk rules

Spoof analysis

Photo, replay, mask, deepfake and video-stream substitution are analyzed separately from face-to-document comparison. Results feed the verdict before product access — as a risk signal, not an absolute guarantee.

Step 3 · face and video spoofLayer passed

Photo instead of facenot found

Recorded videonot found

Masknot found

Deepfake signsnot found

Video stream substitutionnot found

Layer result — in verdictSeparate from face-to-document match

Device signals

The verdict includes not only face and document but session environment: runtime, network, model and OS. VPN or proxy alone does not mean decline — it is context alongside other layers.

Step 4 · Device IntelligenceContext collected

Simulatorno

Emulatorno

Rootno

VPNcontext

Proxy / TORno

Model and OSknown

VPN — context, not a verdict

Session signals addedSame flow, no separate anti-fraud form

Verdict

Results return to your product flow. You set the rules: approve, decline or send for manual review — instead of automatically rejecting honest users on ambiguous results.

Step 5 · session verdictReady

OverviewLivenessFace matchEvents

14:02Document captured

14:02Liveness confirmedok

14:03Face matchmatched

14:03Video spoofnot found

14:03Device signals accounted for

Approverule A

Reviewrule B

Declinerule C

Responseto your flow

You make the decisionWe return layer results and final status

Anti-fraud starts at registration

The right face on a suspicious device is not yet a safe customer

Device Intelligence evaluates onboarding context: simulators, emulators, root, VPN, proxy, TOR, model and OS. The product sees not only who is on camera but where the session comes from — before a wallet, limit or other service opens.

VPN or proxy alone is not a verdict: device signal is evaluated together with face, document and spoof layer. A burst of sign-ups from one environment is visible by device; repeat faces — via Face Search 1:N. Both layers feed one verdict so account farms do not look like unrelated applications.

Overview · Device IntelligenceRisk

Face passed checksDocument, liveness and match — no issues

Emulatordetected

Rootdetected

VPN / proxycontext

TORcontext

Device modelknown

OSknown

!Registration — manual reviewAction defined by your risk rules

Repeat sign-ups

A returning face is visible before a new account opens

Face Search 1:N compares the new face against your connected database of verified users. A potential match becomes a verdict signal before product access. Average face-database search speed is about 1.5 seconds.

Face Search · 1:NSearch complete

New registrationFace compared against connected database1:N

match

Average face database search speed~1.5 sec

!Potential face repeatSignal enters verdict before new account opens

For developers

Classic integration

If your team prefers to assemble architecture manually, connect ready SDKs and API.

JS · Webview

Web SDK / Flow Widget

Ready JS camera component for web registration or in-app Webview. Verification runs inside your onboarding; design adapts to your brand.

JavaScriptCustom designWebview

iOS · Android

Mobile SDK

Native verification flow in mobile apps: camera access and biometric micro-packets even on weak connections.

SwiftKotlinOffline-ready

Backend

REST API

Layer results and final verdict arrive in your backend logic: risk rules and decisions stay on your side. X-API-KEY auth.

RESTX-API-KEYServer-side

AI · AGENTS

MCP server for AI agents

Connect verification to agent scenarios without writing code. Works with Claude, GPT and any MCP client.

MCPClaudeAI-agents

fintech_onboarding — REST API

# Customer onboarding / fintech_onboarding

POST /v1/verification/check/

Host: api.biometric.vision

X-API-KEY: sk_live_••••••••••••••

Content-Type: application/json

{ "scenario": "fintech_onboarding", "checks": ["document", "liveness", "face_match", "device"] }

# → 200 OK

{ "status": "completed", "live": true, "face_match": "match", "device": { "vpn": true } }

<1 seccheck time

99.1%accuracy

37 / 570+global ranking

180+countries supported

Questions and answers

A dedicated layer analyzes photo, replay, mask, deepfake and video-stream substitution signs. When triggered, the result feeds the overall verdict and your rules can decline registration or send it for manual review. You can test your own samples before contract.

500 checks a month — free

Check risk before opening the product — without extra loops for the customer.

Connect document, liveness and face checks, spoof analysis and device signals to your onboarding. Start in the cabinet or request a demo for your product flow.

Start free Request a demo

Numbers and facts

433+: Document types —Passports and hundreds of other document types worldwide
180+: Supported countries —Global coverage for verification across jurisdictions
40 sec: Onboarding time —User registration and verification in under a minute
99%: Technology accuracy —High-precision AI document recognition and verification

Honest customers move fast. Face, video and device spoofing risks are visible before opening the product.

Fintech grows fast. Fraud walks through the same door.

From document to verdict — one scenario without extra forms

Document

Liveness and face match

Spoof analysis

Device signals

Verdict

The right face on a suspicious device is not yet a safe customer

A returning face is visible before a new account opens

Classic integration

Web SDK / Flow Widget

Mobile SDK

REST API

MCP server for AI agents

Questions and answers

How does verification handle deepfakes and video spoofing?

Will extra checks hurt conversion?

What if the device looks suspicious?

Can you detect multi-accounts?

Is AML mandatory in every scenario?

Can we skip full KYC again in another product?

What does the user see on a weak device or bad capture?

Where is data stored?

Where to start?

Check risk before opening the product — without extra loops for the customer.