Guide · Technology
How AI face reading works
On-device & private · Updated July 2026
AI face reading works by turning a photo into numbers, then reading those numbers. In FaceTale, an on-device AI model maps a 478-point face mesh onto your face right in your browser, derives dozens of geometric measurements from it, and matches them to a fixed set of personality archetypes. Your original photo never leaves your device — only anonymous geometry is used.
Step 1 — an on-device face mesh
When you upload or capture a photo, a face-mesh model (MediaPipe) runs entirely in your browser and locates 478 landmark points across your features. Because it runs on your device, your photo never reaches our servers and is never stored.
Step 2 — geometric measurements
From those points, FaceTale computes dozens of proportions, angles, and symmetry values — the width of the nose, the arch of the brows, the spacing and openness of the eyes, jawline angle, and more. These are proportions only; they describe geometry, not identity.
Step 3 — a deterministic archetype match
The measurements map to one of 16 archetypes using a fixed, deterministic method: the same face always produces the same numbers and the same archetype. There is no random guessing — the result is consistent and reproducible.
Step 4 — an AI-written reading
Finally, AI writes the reading around your archetype, weaving in the five features, twelve palaces, and five elements from the tradition of Chinese face reading. The measurement is real technology; the interpretation is a traditional art.
Why on-device matters
Running detection on your device is the core of FaceTale's privacy model: the sensitive part — your actual photo — never travels over the network. That is a deliberate, technical choice, not just a policy.
FaceTale pairs the traditional Chinese art of face reading (面相) with modern AI. Your reading is a cultural interpretation for self-reflection — not medical, diagnostic, or a guaranteed prediction.