Rate my face

The AI loads a 68-point facial landmark model (a TensorFlow.js implementation derived from the iBUG 300-W landmark spec) and locates the corners of your eyes, the nasal tip, the philtrum, the zygion (cheekbone) points, the gonion (jaw angle) points, and the upper and lower lip border. From those landmarks, four distinct measurement passes run in sequence.

Symmetry is computed by mirroring the left half of the face onto the right and measuring landmark displacement at matched points. The score reports millimeter-equivalent deviation normalized by interpupillary distance — the same normalization used in clinical craniofacial papers so the score is independent of how close you are to the camera.

Smile authenticity uses Duchenne markers from Ekman & Friesen's Facial Action Coding System (FACS): AU6 (orbicularis oculi cheek raise) co-occurring with AU12 (zygomaticus major lip corner pull). A polite smile activates AU12 alone; a genuine smile activates both. The AI measures crow's-feet bunching above and below the lateral canthus and compares it to lip corner elevation.

Eye warmth measures palpebral aperture narrowing and lower eyelid bunching — the eye-side correlates of a real smile. Expression score aggregates cheek raise, mouth corner curvature, brow tension, and overall facial relaxation into a composite. The four sub-scores are weighted (smile authenticity is the heaviest because research finds it dominates first impressions) and mapped to a 0–100 rating that approximately corresponds to a 1–10 face rating scale.

A face rating is a single-photo, single-moment snapshot. It is not your "value" or your "ceiling" — it is the strength of the first impression that this specific photo creates. The same person photographed five minutes apart, with the same outfit, in the same room, can score thirty points apart depending on whether they were genuinely smiling, whether the light was on their face or behind it, and whether the camera was below or above eye level.

The most useful interpretation is comparative. If you upload your three best photos and one scores 78 while the others score 52, 56, and 61, the AI is telling you which photo creates the strongest first impression — and the gap between them is almost entirely about expression and lighting, not facial structure. That is the single most valuable signal a face rater can give you, because it is actionable.

Treat the absolute number with healthy skepticism. Treat the ranking between your own photos with serious attention. Use the score to find the photo that consistently leads with your best expression, then use that one for dating apps, LinkedIn, or your professional headshot.

Photofeeler (per their 2026 pricing page) crowdsources human votes at roughly $0.20 per vote, with results returned in hours to days. Useful for absolute attractiveness ratings from real people; not useful when you need a result before posting a profile tonight.

QOVES charges $150/year (per their 2026 FAQ page) for human-reviewed reports. The depth is real, but the unit price gates first-time users.

Aurale (per testimonial copy indexed on aurale.app, 2026) is $49 one-time with a 30-day money-back guarantee. Your photo is uploaded to their server for processing.

Umax uses an Adapty paywall ladder of weekly/monthly subscriptions ($3.99–$24.99 SKU range as of their 2024 ledger). Photo uploaded to server.

RealSmile face rating is free, instant, and 100% client-side — your photo never leaves the browser. The optional QOVES alternative upgrade ladder is one-time pricing: $29 for a Dating Photo Audit, $49 for the Photo Ranker, $99 for the full Glow-Up Plan. No subscription. No re-billing. See the academic citations behind our scoring.