< 400ms
Median estimation latency
Age Estimation
Estimate age. Skip the document.
A single passive selfie returns an age estimate, calibrated confidence, and a clear pass/fail against your threshold. Step up to documents only when it matters.
0
Documents required
24/7
No queue, no reviewer
Capabilities
What's in the box.
Production-ready building blocks. Wire them up the way your team needs.
Single-frame estimation
One selfie, no prompts. We return an age band, point estimate, and calibrated confidence in under half a second.
Adjustable thresholds
Pick the buffer that fits your jurisdiction (18, 21, 25) and we tune the decision boundary against your risk appetite.
Step-up to documents
When estimation falls inside the uncertainty band, automatically promote the user into a full document check.
Privacy by default
Frames are processed in-memory and discarded after scoring. No biometric retention unless you opt in.
How it works
Plug it in. Ship in days.
01
Capture
User aims the camera at their face. SDK selects the cleanest frame automatically.
02
Score
Our model returns a point estimate and uncertainty band in the response payload.
03
Decide
Compare against your configured threshold and act, or step up to a document check.
04
Audit
Every decision is logged with model version, threshold, and signals for regulator review.
Use cases
Built for teams that move.
- Adult content gating where document upload would harm conversion
- Age-restricted retail (alcohol, tobacco, vape) checkouts
- Social platform sign-up before unlocking riskier features
- Gambling and gaming session gating
- Streaming and creator platforms with age-rated content
FAQ
Common questions.
How accurate is age estimation versus a document?
Document checks remain the gold standard for legal age proof. Age estimation is calibrated for low-risk flows and best paired with a step-up to documents for borderline cases.
Is this compliant with the UK Online Safety Act?
Age estimation is recognised by Ofcom as a "highly effective" age assurance method when calibrated correctly. We provide the audit trail and accuracy reporting regulators expect.
Can the model run on-device?
We ship a lightweight on-device variant for capture quality and pre-screening. The full-accuracy model runs server-side with a sub-400ms median.
Get started
Ready to ship age estimation?
Tell us your use case and we'll design the flow around it, then hand you the SDK and a sandbox to try it.