Goldeneye

面相 · on-device reading engine: booting…

看看你的面相 · 实时或照片,仅供娱乐 本地处理 · on-device

FACE LOCK : SEARCHING
FPS  · LMK 0
无法使用摄像头 · Camera unavailable
请验证你的邮箱 · Please verify your email
选项 Options
Reading
Capture
AI
key in .env · server-side only

For entertainment & cultural interest only. 面相 is a traditional folk practice, not science — this app makes no real claims about anyone's character, health, or destiny. Frames are processed locally in your browser; nothing is uploaded (photos are read on-device too). Camera needs localhost or HTTPS; photo upload works anywhere. The 直读 / Candid tone surfaces traditional cautions as well as strengths — still folklore, still not judgment of any real person. 痣相 / Moles are detected on-device (dark spots at classical positions) and read for their traditional meaning — approximate and for fun; markers show on uploaded photos. 气色 / Qì-Sè reads the live complexion color at palace zones (relative to your skin baseline) for a dynamic 气运 hint that changes each capture — noisy and lighting-dependent, strictly entertainment. Any health-adjacent note is gentle 养生 only; never a diagnosis — see a doctor for real concerns. 屏幕补光 / Flash briefly turns the whole screen white when you Read Face, using the display as a fill light to even out 气色 in dim rooms (best on a bright laptop/Mac screen). Live-camera only — it can't re-light an already-loaded photo, so it does nothing in photo mode. Single gentle flash per capture (no strobe). 猜年龄 / Guess estimates age from facial proportions + skin-texture (wrinkle) cues to pre-fill 流年 — a rough guess, often off by years; always editable. 神采 / Shén is a dynamic proxy for 「神」 computed from eye openness, symmetry, and (if iris landmarks are present) gaze centering — now multi-frame smoothed with blink rejection so a blink no longer drops it. 门派 冰鉴 (曾国藩, 一身精神具乎两目) and 水镜 lean on it. Honest caveat: this is a crude, blink-sensitive, single-frame approximation of 神, not a real read of someone's spirit or a judgment of their state — strictly for fun. AI 增强 is optional: when enabled, the local server (key from .env, never exposed to the browser) sends the computed reading and scores — never the photo — to your provider for a richer write-up. Leave it off, or just open the static file, and everything runs fully on-device.

HUD // Readout

awaiting capture
Press Read Face. The reading spoken into the wearer's ear will appear here…