Your face score, free in 30 seconds.
Full 17-metric report, $14.99.

RealSmile is an AI face analysis app that scores your photo on 17 facial-geometry metrics, returns a 100-point geometry-conformance score (proximity to ideal ranges, not a population rank), and emails a written PDF with a 30-day improvement plan in under 2 minutes. The free tier returns your score and your single strongest metric. Paid unlocks every metric, your validated Impression Percentile, and the personalized PDF. The scan needs no subscription. Your photo is processed and deleted right away. Nothing is kept unless you choose to save it.

RealSmile's Face Score is validated against human judgment: on 1,566 held-out faces from the SCUT-FBP5500 benchmark it correlates with the averaged ratings of 60 people at r ≈ 0.8 (5-fold cross-validated). The geometry scores are computed from 68 facial landmarks using ratios from peer-reviewed attractiveness research (Rhodes 2006, Penton-Voak and Perrett 2001). A score reads the photo as well as the face — across 1,350 people who scored two or more photos of themselves, the same face's Face Score spanned a median of 13 percentile points between photos, so lighting, angle and expression are worth getting right. The full study is published at /face-score-validation.

The free score and 17-metric breakdown return in 60 seconds. The personalized PDF report generates inside 2 minutes and a copy emails automatically. No queue, no scheduling, no human review delay. For visible appearance change after acting on the plan, fixable factors (lighting, angle, expression, hairstyle, grooming, skin clarity) respond inside 30 days; bone-structure factors do not change without intervention.

The engine calculates 17 facial-geometry measurements: canthal tilt, orbital tilt symmetry, eye shape (hunter eye index), eye spacing, brow arch, brow-eye proximity, facial symmetry, facial thirds balance, facial proportion balance, facial width-to-height ratio (FWHR), midface ratio, jaw taper, jawline angle, chin proportion, nose proportion, philtrum ratio and lip ratio. Jawline angle and eye spacing are withheld when a front-on photo cannot measure them reliably, so a typical report shows 15. Each measurement is taken from your 68 facial landmarks and mapped against a gender-calibrated reference range, with research citations attached. Alongside the measurements you get your Face Score — a separately validated 1-99 percentile of how people tend to rate your photo.

The scan and your validated Face Score (percentile) are free — 1 scan a day with no signup, 5 with a free account. The full 17-metric report with PDF and 30-day plan is $14.99, one-time. The Dating Photo Audit ($19.99) and the Professional LinkedIn Audit ($49) are one-time too. The only recurring plan is Pro at $3.99/week, and it is optional. See full pricing at /pricing.

Not by default. On desktop, analysis runs in your browser. On mobile, your photo is processed in memory by our scan server and deleted immediately after the landmarks are measured — never written to disk. By default nothing is kept: no photo, no copy, nothing used to train anything. There are three exceptions and all three are your choice — you can save a scan photo to your own account, you can submit a photo to the community rating pool, and the paid Audit keeps a thumbnail so you can reopen your report. Each one is deletable, and deleting removes the file itself, not just the link to it. To delete your scores too, email hello@realsmile.online with the word "delete" and the records are purged the same day.

RealSmile is one-time payment for an instant 17-metric AI report. Human-vote services like Photofeeler cost $20 to $100 per photo test with no metric breakdown and a wait queue — RealSmile also offers a free Human Score from real 1-10 votes (see /photofeeler-alternative). Qoves charges $150 per year for ongoing human consults. FaceApp is a filter and aging toy, not a scoring system. Same landmark math the academic literature uses, no subscription, no scheduling, and the PDF is yours to keep.

The free version returns your overall geometry-conformance score out of 100 and your single strongest metric. The full report adds all 17 metrics with tier labels, your validated Impression Percentile, a 7-9 page PDF, your facial archetype, and a personalized 30-day plan ordered by points-recoverable per hour of effort. The free tier exists to let you sanity-check the system before paying anything.

Softmaxxing is completely safe and is the only path RealSmile recommends. Softmaxxing covers grooming, skincare, fitness, posture, hairstyle, expression coaching, and photo conditions, all of which respond to behavior change inside 30 days with no medical risk. RealSmile does not recommend or score any surgical procedure. The 30-day plan ranks fixes by leverage so the safest, highest-impact changes happen first.

AI can measure the objective geometric signals that correlate with rater agreement, but it cannot tell you whether you are attractive to a specific person. RealSmile reports geometric ratios (symmetry, FWHR, canthal tilt, facial thirds) and a Face Score that correlates with averaged human ratings at r ≈ 0.8 on 1,566 held-out faces. The photo moves the numbers — the same face's Face Score spans a median of 13 percentile points between photos (1,350 people who scored two or more) — so use them for photo optimization, not as a verdict on your worth.

Yes. All 17 metrics apply to any face. Benchmarks are computed separately by gender so a female score is compared against female reference distributions, not male ones. The 30-day plan adjusts for gender-typical photo conventions on dating apps and LinkedIn, including lighting, angle, expression coaching, and wardrobe notes.

The geometric metrics draw on Rhodes 2006 (Annual Review of Psychology, PMID 16318594) on cross-cultural attractiveness preferences, Penton-Voak and Perrett 2001 on symmetry and sexual dimorphism, and the Geniole, Carre, and McCormick FWHR meta-analysis (PMID 26181579) on facial-width cues. Expression scoring uses Ekman FACS coding for Duchenne markers. First-impression weighting is calibrated to Willis and Todorov 2006 (PMID 16866745).