Identity studio

Only images marked ok and left checked are aggregated. Uncheck any image whose overlay looks misaligned before computing.

Optional, experimental: enroll your identity profile as a face-unlock template, and wallet actions below will first ask you to look at the camera (plus two quick liveness checks in random order — a head turn and an eye closure). The template is ratio statistics and, when the on-device embedding model is available, one face-embedding vector — no photos, no landmarks — stored in this browser and never uploaded. This is a convenience gate, not a security boundary: your wallet's own credential (passkey or extension) is still required, and a face can be spoofed by a determined attacker.

Text-to-image via Visbir's backend: the identity constraints from the Identity Bible are applied server-side, and your profile's measured ratios can be embedded in the prompt. Only text ever leaves the browser — never photos or landmarks. Each result is then landmarked and scored client-side against your profile, so "does this look like me?" becomes a number. Prompt-based generation approximates the written constraints; it cannot clone a face — the score tells you how close each candidate landed.