Selfie age estimation
Collect a live selfie
Capture a live selfie from your users through a guided, mobile-friendly experience built to minimize drop-off.

Analyze age and detect spoofs
Estimate age with a purpose-built model trained on more than a million ethically sourced images. Confirm each submission comes from a real, live person to detect presentation and deepfake attacks.

Automate decisions and redaction
Automatically compare results to your age thresholds and route borderline users to fallback methods. Automate data redaction to meet relevant privacy regulations.

Key features
Highly-performant age estimation with powerful controls
Configurable age thresholds and buffersToggle description visibility
Set custom age cutoffs and safety buffers to meet compliance requirements.
Frequently updated age estimation modelsToggle description visibility
Persona's models are regularly retrained on ethically sourced data to stay accurate across diverse populations.
Equitable across demographics and devicesToggle description visibility
No instances of material bias across sex, skin tone, and age, with accessible UX on all modern operating systems and devices.
Support for immediate redactionToggle description visibility
Automatically delete data as soon as estimation is complete, supporting privacy-by-design and regulatory compliance requirements.
Multi-frame analysisToggle description visibility
Analyze multiple selfie frames to catch spoofs and manipulation attempts that single-frame checks might miss.
Auto-captureToggle description visibility
Automatically capture photos at the ideal moment to decrease user error and boost conversion rates.
What sets Persona’s selfie age estimation apart
Persona analyzes risk signals throughout the age assurance process to flag signs of manipulation or spoofing attacks.
With a mean absolute error of under 2 years, Persona’s age estimation models rank among the most accurate in the industry.
Start with a lightweight check and only step up to more robust methods when needed, keeping friction for most users low.