A real person, every time.
Passive liveness detection from a single selfie frame. No blinks, no head-turns, no prompts — just silent presentation-attack and deepfake detection that beats printed photos, screen replays, masks and injected video.
One call. Every check.
Each capability runs in parallel and returns in a single structured response.
Single-frame analysis
Liveness is decided from one ordinary selfie — no actions to perform. Users tap once and move on, which is why pass-through stays high.
Presentation-attack detection
Detects printed photos, cut-outs, screen replays and 3D masks by analysing texture, micro-reflection and depth cues invisible to the eye.
Deepfake & injection defence
Flags virtual cameras, emulators and injected video — the channel-level attacks that bypass camera-only liveness checks.
No head-turns
No blink, smile or turn-your-head choreography. Accessible by default and far less likely to be abandoned mid-flow.
ISO 30107-3 conformant
Independently lab-tested against the ISO/IEC 30107-3 presentation-attack standard, with reports you can share with auditors.
Any camera
Runs on front cameras across smartphones, tablets and webcams without special hardware, depth sensors or IR.
Every spoof, one model.
Presentation attacks and injection attacks are scored together so you do not have to chain separate checks. The model is tuned to hold a high genuine pass-rate while still catching the long tail of sophisticated spoofs.
Friction is the enemy of completion.
Active liveness asks users to blink, smile or rotate their head — steps that confuse, exclude and lose customers. Passive liveness needs none of that, which lifts pass-through and works for users who cannot follow motion prompts.
- One tap — no blink, smile or head-turn choreography
- Accessible to users who cannot follow motion prompts
- 99.2% genuine pass-through keeps onboarding completing
- Retry hint on poor capture instead of a hard reject
A single liveness call.
Send one selfie frame and receive a typed result with a liveness verdict, a spoof score and the specific attack types detected. Combine it with Face Matching in one session to bind liveness to identity.
const r = await othento.face.liveness({
selfie: selfieImage,
});
// r.isLive -> true
// r.spoofScore -> 0.02
// r.attacksDetected -> []
// r.standard -> "ISO_30107-3"Four steps, about a second.
Capture
A single selfie is captured through the SDK — no actions, prompts or movement required.
Analyse
The model reads texture, micro-reflection and depth cues from the one frame.
Detect spoofs
Presentation and injection attacks are scored — photos, screens, masks and deepfakes.
Confirm
A live / spoof verdict with a spoof score and attack types is returned and logged.
Onboarding, everywhere.
Remote onboarding
Guarantee a live applicant before opening an account, without adding steps that lose them.
Step-up authentication
A frictionless liveness check before high-risk actions like adding a payee or raising a limit.
Account recovery
Stop SIM-swap and account-takeover attacks by proving a real person at recovery time.
Manual review can't keep up.
| Othento | Manual review | Legacy vendor | |
|---|---|---|---|
| User effort | Single selfie | Video call | Blink / turn / smile |
| Genuine pass-through | 99.2% | Reviewer-limited | Drops with friction |
| Deepfake / injection | Detected | Hard to spot | Often missed |
| ISO 30107-3 testing | Lab-certified | N/A | Sometimes |
| Special hardware | None — any camera | N/A | IR / depth sensor |
| Time to decision | Under a second | Minutes | Several seconds |
Switching to Othento passive liveness removed the blink-and-turn steps our users hated. Completion went up and we have not seen a single replay or deepfake get through since.
Questions, answered.
No. Liveness is decided from a single selfie frame with no actions or prompts. That passive approach keeps onboarding fast, works for users who cannot follow motion instructions, and removes the most common cause of drop-off.