Capability

Liveness detection API

Document capture, liveness, face match, mDL, and jurisdictional age-tier resolution. Signed webhook returns an eligibility signal.

Talk to complianceView docs

Capabilities

Passive liveness scored against the spoof types attackers actually use

Passive liveness

Single-frame liveness assessment with no UX prompts — no head-turn, no smile, no read-the-prompt. Sub-second result; the user does not know liveness is being scored. Same conversion path as a non-liveness flow.

Spoof coverage

Defeats printed photos, screen replays, video injection, deepfake injection, and 3D-mask attacks. Continuous model retraining against new spoof samples; threshold tuned per merchant flow if the default false-rejection rate isn't a fit.

Liveness as part of the flow

Same SDK, same signed webhook. Liveness scoring runs server-side after capture; the merchant receives only the eligibility decision, not the raw liveness score. Slot it into age-verification, identity-verification, or any custom flow.

How it works

Liveness inline with verification — no extra round trip

1. Capture

Selfie captured in the same session as the document

2. Score

Passive liveness model runs server-side; spoof signals flagged

3. Signal

Eligibility decision returned via signed webhook

Liveness detection is rarely the whole flow — it sits inside age verification, identity verification, or a high-trust onboarding step. Most buyers reach this page from one of those. Related coverage:

Age verification — age-tier flow with liveness scoring inline.

Identity verification — KYC and onboarding flows with liveness scoring inline.

Regulated commerce — gambling, alcohol DTC, adult content, and cannabis flows where liveness scoring is standard.

US state coverage — which state laws reference biometric handling and how Stile retains liveness signals.

FAQ

Liveness, the parts buyers ask about

Printed photos, on-screen replays from phone or laptop, video injection (a real video stream substituted for the camera feed), deepfake injection (a synthesized face stream), and 3D-mask attacks. The model is retrained against new spoof samples on a rolling basis; new attack types are tracked in the changelog.

Get started

Talk to compliance

Tell us the spoof types you have seen in production and the false-rejection budget your flow tolerates. We will recommend a threshold and a model variant.

Bring liveness scoring to your verification flow

Email goes to alex@stile.id directly. Engineering and compliance are CC'd on the reply.

Email alex@stile.id