Video KYC (V-KYC)¶

Definition¶

Video KYC (V-KYC) is a method of customer identity verification conducted through a live, real-time video call between the customer and a trained verification agent. It combines the assurance of in-person verification with the convenience of remote onboarding, enhanced by AI-assisted checks for face matching, liveness, and document reading.

How Video KYC Works¶

sequenceDiagram
    participant C as Customer
    participant App as Mobile/Web App
    participant Q as Queue System
    participant A as V-KYC Agent
    participant AI as AI System
    participant BE as Backend

    C->>App: Initiates V-KYC session
    App->>Q: Join video queue
    Q->>A: Assign next available agent
    A->>C: Video call begins (recorded)

    A->>C: Ask to show front of ID
    C->>A: Holds up ID document
    AI->>AI: OCR + document forensics
    AI->>AI: Face detected on document

    A->>C: Ask to show back of ID
    C->>A: Flips document
    AI->>AI: Extract additional data

    A->>C: Ask to look at camera
    AI->>AI: Face liveness check
    AI->>AI: Face match (selfie vs ID photo)

    A->>C: Verification questions
    C->>A: Answers questions

    A->>AI: Request Aadhaar OTP
    AI->>C: OTP sent to registered mobile
    C->>A: Reads OTP aloud or enters

    A->>BE: Submit verification with AI scores
    BE->>BE: Risk assessment
    BE->>C: KYC approved/rejected

    Note over App,BE: Entire session recorded with<br/>timestamp, geo-location, and audit trail

V-KYC Regulatory Framework¶

India (RBI) — Most Detailed V-KYC Guidelines¶

RBI's January 2020 amendment to the KYC Master Direction formalized V-KYC:

Requirement	Details
Agent qualification	Must be official of the Regulated Entity (bank/NBFC), trained in KYC/AML
Video quality	Clear, real-time, two-way audio-video
Recording	Entire session must be recorded and stored
Geo-tagging	Customer's live GPS location captured
Document capture	Customer must display original OVDs — high-res capture
Aadhaar OTP	Must be verified during the session
PAN verification	Cross-checked in real-time
Face match	Live face matched with photo on ID (AI-assisted)
Randomized checks	Agent asks questions to verify identity awareness
Concurrent sessions	Agent cannot handle multiple V-KYC calls simultaneously
Audit trail	Complete log with timestamps, agent ID, customer ID
Failure handling	If video drops, session must restart from beginning
Redaction	Aadhaar number must be masked/redacted in stored recordings

Germany (BaFin VideoIdent)¶

Requirement	Details
Agent	Trained identification agent (can be third-party)
Encryption	End-to-end encrypted video stream
Document check	Agent performs visual security feature inspection
Hologram check	Customer must tilt document to show holographic features
Photo match	Agent visually confirms face matches document
Random questions	From ID data to verify customer knows their own details
Recording	Session recorded, stored for 5 years

Other Jurisdictions¶

Country	V-KYC Status
France	PVID (Remote Identity Verification) framework — ANSSI certified
Italy	SPID video identification accepted
Spain	SEPBLAC guidance allows video-based identification
Austria	VideoIdent accepted under BWG
India	Most comprehensive V-KYC framework globally

V-KYC Technology Stack¶

graph TD
    subgraph "Client Side"
        A[WebRTC Video Stream]
        B[Camera Controls]
        C[Location Services]
        D[Screen Sharing blocked]
    end

    subgraph "AI Layer (Real-Time)"
        E[Face Detection]
        F[Face Liveness]
        G[Face Matching]
        H[Document OCR]
        I[Document Forensics]
        J[Quality Assessment]
    end

    subgraph "Agent Interface"
        K[Video feed + AI overlays]
        L[Extracted data display]
        M[Verification checklist]
        N[Decision controls]
    end

    subgraph "Backend"
        O[Video recording + storage]
        P[Database verification APIs]
        Q[Risk scoring]
        R[Audit trail]
    end

    A --> E & F & G & H & I & J
    A --> K
    E & F & G & H & I & J --> K & L
    K & L & M --> N
    N --> Q
    A --> O
    P --> L
    Q --> R

    style K fill:#4051B5,color:#fff

Key Technology Components¶

Component	Technology	Purpose
Video streaming	WebRTC	Low-latency, peer-to-peer video
Recording	Server-side recording	Compliance — entire session stored
Face detection	SCRFD / BlazeFace	Real-time face tracking during call
Face liveness	Passive liveness model	Ensure customer is live (not replay)
Face matching	ArcFace / AdaFace	Compare live face with ID document photo
Document OCR	LayoutLMv3 / PaddleOCR	Extract data from documents shown on camera
Document forensics	Forensic CNN	Check document authenticity in real-time
Geo-location	GPS + IP geolocation	Capture customer location
Encryption	TLS 1.3 + E2E encryption	Secure video transmission
Queue management	Custom or vendor (Twilio, Vonage)	Route customers to available agents

V-KYC vs Other eKYC Methods¶

Aspect	Automated eKYC	Video KYC	In-Person KYC
Human involvement	None (AI only)	Agent + AI hybrid	Fully human
Assurance level	High	Very High	Very High
Speed	30-60 seconds	5-15 minutes	30-60 minutes
Cost per verification	$0.50-$5	$5-$15	$15-$25
Scalability	Very high	Limited by agents	Limited by branches
Availability	24/7	Agent working hours	Branch hours
Spoofing resistance	Model-dependent	High (human + AI)	Very high
User experience	Best	Good (wait times)	Worst
Regulatory acceptance	Growing	High	Universal

When to Use Video KYC¶

graph TD
    A[Verification Request] --> B{Risk Level}
    B -->|Low risk| C[Automated eKYC<br/>Document + Selfie]
    B -->|Medium risk + automated failed| D[Video KYC]
    B -->|High risk| D
    B -->|Regulatory requirement| D
    B -->|Customer preference| D

    D --> E{V-KYC Outcome}
    E -->|Approved| F[Account Opened]
    E -->|Inconclusive| G[Escalate to Senior]
    E -->|Rejected| H[Rejection]

    style C fill:#2E7D32,color:#fff
    style D fill:#1565C0,color:#fff

Typical V-KYC use cases:

Use Case	Why V-KYC
Automated eKYC fallback	Face match or liveness failed — human verification as escalation
High-value accounts	Private banking, large deposit accounts require higher assurance
PEP customers	Additional human verification layer for politically exposed persons
Document edge cases	Damaged, unusual, or hard-to-read documents
Regulatory mandate	Some products require face-to-face equivalent verification
Customer preference	Some customers prefer human interaction
Elderly users	Users who struggle with automated selfie/liveness flow

V-KYC Agent Training Requirements¶

Training Area	Topics
KYC/AML basics	CDD, EDD, risk assessment, SAR identification
Document recognition	Identifying genuine documents, security features
Fraud detection	Common impersonation tactics, social engineering
Technology	Using the V-KYC platform, AI overlay interpretation
Communication	Professional video call conduct, customer guidance
Privacy	Data protection obligations, handling sensitive data
Escalation	When and how to escalate suspicious cases
Regulatory	Specific V-KYC regulatory requirements

V-KYC Challenges¶

Challenge	Details	Mitigation
Wait times	Customers wait for available agent	Hybrid: try automated first, V-KYC fallback
Bandwidth	Poor network causes dropped calls	Adaptive bitrate, reconnection logic
Cost	$5-15 per verification (agent + infrastructure)	Reserve for high-risk / fallback only
Scale	Linear scaling — more calls need more agents	Outsource V-KYC to specialized providers
Agent fatigue	Repetitive work causes quality drop	Rotation, breaks, AI-assisted checklists
Language	Multi-lingual customer base	Multi-lingual agents or translation AI
Spoofing	Deepfake during video call	Real-time liveness + random challenges

Key Takeaways¶

Summary

V-KYC combines human judgment with AI assistance for the highest remote verification assurance
India has the most comprehensive V-KYC framework — mandatory recording, geo-tagging, agent qualification
Best used as a fallback or escalation path — not primary verification (too expensive to scale)
Cost is $5-15 per verification — 3-10x more than automated eKYC
Deepfakes pose a growing threat even for V-KYC — real-time face swap can fool human agents
V-KYC agents need specialized training in documents, fraud, regulation, and technology