PhotoCheck API

Photo checks for identity verification and onboarding

Before a selfie goes to your identity provider, PhotoCheck can confirm that it contains exactly one face, large enough to use, and that it doesn't look AI-generated. That catches unusable uploads early and flags generated faces cheaply, before an expensive verification call. It is a pre-check, not a liveness or identity check.

The problem

Identity checks are priced per attempt, and many attempts fail for dull reasons: no face, two faces, a face too small to match. Others are deliberate, with a generated face submitted in place of a real one. Both are cheaper to stop before the vendor call.

The checks that matter

Exactly one faceRejects group shots, empty frames and photos of a screen with several faces.
Face sizeThe box tells you if the face fills enough of the frame to be usable.
AI scoreFlags generated faces before they reach your identity vendor.

A decision rule

Built on POST /api/v1/analyze. Thresholds are starting points; tune them to what a mistake costs you.

JavaScript
const r = await check(selfie);
if (r.faces.count !== 1) return "retake:one_face";
const f = r.faces.faces[0].box;
if (f.width < r.image.width * 0.25) return "retake:closer";
if (r.ai.score >= 0.5) return "manual_review";
return "send_to_idv";

API reference, Node.js quickstart, pricing.

Questions

Is this a liveness check?

No. PhotoCheck looks at one still image. It can't tell a live person from a printed photo or a screen. Use your identity provider for that.

Does it compare the selfie to an ID?

No. PhotoCheck detects faces; it never recognises or matches them.

Other use cases