Test the live session itself.
Detect generated video, face swaps, virtual-camera injection and replay, with face and voice deduplication across identities.
Early liveness checks were built for printed photographs and screen replay. Generated video and virtual cameras require evidence from the full session, not one sampled frame.
Liveness cannot be reduced to whether a face moved on command.
- Capture pathNative camera
- Frame continuityReview required
- Face reuseNo cross-identity match
- Session verdictInsufficient evidence
What the module returns
The output is structured for a decision engine and readable by the person who must review it later.
Session integrity verdict
The decision separates a live capture from replay, virtual-camera injection and generated media evidence.
Face vector reuse
A face matched across distinct identities is surfaced with the number of linked records and session references.
Voice consistency
Audio evidence is checked for replay characteristics and repeated voice vectors across applicants.
Review packet
Analysts receive reason codes tied to session moments rather than an unexplained pass or fail.
The recording stays where regulation requires.
Raw V-CIP recordings and derived session evidence remain on systems located in India and can stay inside the institution's environment.
Capture begins
The institution's approved V-CIP surface collects the session and device telemetry.
Media path checked
The service tests capture provenance, frame continuity and signs of an injected virtual source.
Vectors compared
Face and voice representations are deduplicated against identities the institution is permitted to compare.
Evidence retained
The verdict and reason codes join the institution's V-CIP audit record under its retention policy.
Storage location is part of the architecture.
RBI KYC Master Direction requires V-CIP data and recordings to remain on systems in india. The module can run within that boundary.
Test V-CIP Integrity on historical applications.
We reproduce the signals available at the original decision time, then compare the ranked result with later confirmed outcomes.
Book a backtest