bynn-age-verification

Age verification

Two shapes. Pick by threat model.

• Selfie check. One image in, an age estimate out. Fast, no session. Right for a soft age gate where the cost of a wrong answer is low.
• Liveness flow. The image is captured from a live camera under challenge, so an uploaded photo of someone older does not pass. Right for anything regulated.

Base: https://api.bynn.com/v1. Private key.

Selfie check

curl -X POST https://api.bynn.com/v1/age_verification/selfie \
  -H "Authorization: Bearer <YOUR_PRIVATE_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "image_url": "https://example.com/face.jpg",
    "age_verification_model": "age_verification_18_years"
  }'

Send image_url or image_base64_strict, not both. Base64 must be strict, with no line breaks and no data: prefix.

Picking a model

age_verification_model selects the threshold the model was trained on, from age_verification_12_years through age_verification_21_years in one-year steps (12, 13, 14, 15, 16, 17, 18, 19, 20, 21).

Choose the threshold you actually need to enforce. A model trained at the boundary you care about beats reading a raw age estimate and comparing it yourself, because the decision boundary is where the model is calibrated.

legal_age_21 is a separate flag. It sets the legal adult threshold for your jurisdiction, defaulting to 18, and it is what drives is_adult. It is not the same as the model threshold, and setting one does not set the other.

Reading the result

Field	Meaning
verification_status	pass, underage, unknown, pending, or error. Branch on this.
is_adult	Boolean against the legal age threshold.
confidence	high, moderate, or low.
age_estimated	Single best estimate. A prediction, not a fact.
age_range	low and high bounds around the estimate.
face	Detected face attributes.
liveness	status and confidence for whether a real human was in frame.
image_saved, biometrics_saved	Whether anything was retained.
data_destroyed_at	When the data is or will be destroyed.