What RealSmile actually is
RealSmile is a measuring instrument. Point it at a front-facing photo and it finds 68 landmark points on your face. Then it computes 17 measurements from the geometry — canthal tilt in degrees (the tilt of the line from the inner to the outer eye corner), facial width-to-height as a ratio, symmetry as a percentage. Real units you can check, not a vibe dressed up as a number.
One thing matters more than anything else on this page: every number we show is a read on a photo, not a verdict on a person. Change the lighting, the angle, or the lens distance and some of your numbers move — because the photo changed, not your face. That's why the scan comes with a retake coach instead of pretending one snapshot is the truth about you.
We're also strict about the word "validated." The 17 geometry readings are honest measurements, but a measurement is not a rank. The only number we call validated is Your Face Score — free with every scan. We tested it against 1,566 human-rated faces and published how well it tracks what people actually perceive, so you can judge the claim yourself. Everything else is labeled for exactly what it is.
You can read a full sample face report end-to-end before paying for anything, and the price of the paid report is shown before you even upload a photo.
We tell you what can't change
Most tools in this category sell you the same story: everything about your face is a problem, and every problem has a product. We split your results differently — every measurement is marked fixable, partially fixable, or structural. Structural means bone: no routine, no supplement, no 30-day plan moves it, and we say so in plain words.
That costs us money. "Your gonial angle is set by your skull and that's normal, not a flaw" sells fewer reports than "unlock your jawline potential." We write it anyway, because the entire value of a measurement is that you can trust it — and you can't trust an instrument that always finds something to sell you.
When something is fixable, the paid report says exactly what to do about it; when it isn't, the honest answer is the product.
The community keeps it honest
A measuring instrument needs a reference, and ours is people. Real people rate volunteered photos 1–10, anonymously — no comments, no usernames attached, numbers only. Every photo is screened automatically before it enters the pool, and anything flagged is held for a person to review.
The average of 10+ votes becomes a Human Score, and every vote cast makes that reference a little more useful for everyone who comes after.
Contributing is opt-in only, adults only, and reversible: your photo enters the pool when you explicitly say so, and you can delete it whenever you want. Nobody's photo is rated because they ran a scan — scanning and contributing are separate decisions, and the second one is always yours. If you'd rather just rate photos and never show your own, that's a real contribution too.
Our tools
The looksmaxxing test runs all 17 metrics at once. The individual metric pages let you explore each one in depth.
Looksmaxxing Test
Full Analysis17 facial metrics in one analysis: symmetry, canthal tilt, FWHR, jawline angle, hunter eyes, facial thirds, and more. Scored against population averages with a personalized improvement plan.
Smile Analyzer
Genuine vs. fake smile detection based on Duchenne smile research — measures eye muscle engagement vs. mouth-only smiles.
Compare Photos
Upload two photos and see which one scores higher. Useful for picking the best dating app or LinkedIn photo.
Golden Ratio
Facial proportions vs. the golden ratio (1.618). Analyzes facial thirds, eye spacing, and nose-to-mouth ratios.
Attractiveness Test
Beauty score based on facial proportions, symmetry, and expression. Includes what's pulling the score up or down.
Am I Ugly Test
Honest AI analysis of what actually makes a face score high or low — and why you look worse in photos than real life.
Facial Symmetry
Bilateral symmetry comparison across jawline, eyes, brows, and mouth. Measured in percentages, not vibes.
Individual Metrics — also measured in the looksmaxxing test
Jawline
Gonial angle in degrees + definition score
Canthal Tilt
Eye angle — positive tilt is rated most attractive
Hunter Eyes
Lid hooding, tilt, and brow ridge prominence
FWHR
Facial width-to-height ratio — key masculinity metric
Face Shape
Oval, square, round, heart, oblong, or diamond
Masculinity
All 5 key markers of facial masculinity scored
How the measurement works
Landmark detection
A computer-vision model locates 68 key points on your face — eyes, brows, nose, mouth corners, jawline, and chin — from a single photo.
Geometric measurement
From those points we calculate real measurements: symmetry percentages, angle degrees (canthal tilt, gonial angle), distance ratios (FWHR, golden ratio). Math, not opinions.
Honest context
Each reading is placed against population averages and research-backed ranges, and labeled fixable, partial, or structural — so you know what a number means before you react to it.
The landmark model is @vladmandic/face-api, which uses the iBUG 300-W 68-point scheme: a standard map of the eyes, brows, nose, mouth and jaw outline. It is built on TensorFlow.js, a machine-learning library for JavaScript — open tooling you can inspect, not a black box.
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.
The research behind it
Every metric we measure is grounded in published research. Here's the key history.
Da Vinci & Pacioli formalize facial proportions
Luca Pacioli, with illustrations by Leonardo da Vinci, published "De Divina Proportione" — applying the golden ratio (1.618) to human facial proportions. This is the mathematical foundation of our golden ratio analysis.
Duchenne identifies genuine smiles
French neurologist Duchenne de Boulogne showed that genuine smiles activate two muscles: the zygomaticus major (mouth) and orbicularis oculi (eyes). Posed smiles use only the mouth. This is the basis of our smile analyzer.
Ekman & Friesen develop the Facial Action Coding System
Paul Ekman and Wallace Friesen catalogued every human facial movement into Action Units.
FWHR and canthal tilt research emerges
A meta-analysis finds observers judge faces with a wider facial width-to-height ratio (FWHR) as more dominant and more threatening (Geniole et al., 2015). A study of seven ethnic groups finds the eye slant that reads as attractive differs by group (Rhee, Woo & Kwon, 2012). Both became core looksmaxxing metrics.
AI makes it accessible to everyone
Modern AI detects 68 facial landmarks in milliseconds and computes measurements that once required manual analysis by researchers. RealSmile puts this in your browser — free, private, instant.
The data ethic
Four rules, in order. Everything we do with a photo follows from them.
Free scans are ephemeral
Desktop analysis runs entirely in your browser. On mobile, the photo is processed in memory by our scan server and deleted the moment the landmarks are measured — never written to disk. Close the tab and the free scan is gone; no account, login, or email required.
Consent comes first
Nothing about you persists unless you explicitly ask it to. Saving results to a dashboard, volunteering a photo for community rating — each is a separate, clearly-labeled choice. We never quietly convert a scan into stored data.
Deletion propagates
Delete a contributed photo and it leaves the rating pool. Delete your account and your scans, audits, and contributions go with it — deletion means deleted, not hidden.
No training without explicit opt-in
Free-scan photos are never used to train anything, full stop. The only images that can ever improve the scoring model are photos people deliberately contributed under the community consent terms — and withdrawing consent withdraws the photo. Your facial data is never shared with advertisers or data brokers; our analytics count page views, not faces.
Editorial standards
Research-backed claims only
Every metric and correlation we cite references a peer-reviewed study. We link to the original research so you can verify claims yourself. Our research base lists the core studies, each with its authors, year, source and key finding.
Methodology transparency
We use face-api.js (68-point iBUG 300-W model) for landmark detection and publish our measurement methodology. We disclose limitations: 2D photo analysis is inherently less precise than 3D imaging, and scores are relative, not diagnostic.
Regular content updates
Articles are reviewed and updated when new research is published or when tools we review change their features or pricing. Last updated dates are shown on every article.
Honest tool comparisons
When we compare our tools to competitors, we disclose that we built RealSmile. We test competing tools ourselves and report both strengths and weaknesses. We don't fabricate competitor shortcomings.
Who builds this
RealSmile is built and run by one person, Randy — a deliberately small operation, which is exactly why the rules above are written down. One person can't hide behind "policy"; the person who wrote the honesty rules is the person who ships the code, checks the community photos held for review, and answers the support inbox.
It started as a Duchenne-smile analyzer — the genuine-vs-posed smile research was too good to sit in journals — and grew into the 17-measurement scan as people asked for more. The philosophy never changed: established research, real units, and the honest answer even when a flattering one would sell better. Every metric maps to a published source you can check on this page.
How pages here are written and checked
Pages on RealSmile are written with AI assistance and edited by Randy, who owns and runs the site. There is no outside writing team, so one person is accountable for every page.
Links to research papers are checked against PubMed (NCBI) and Crossref, the registries that index published papers: the paper a link opens has to match the authors, year and title named beside it. If a paper can't be confirmed that way, the citation stays as plain text rather than carrying a guessed link.
Several past misattributions, where a page credited a paper with a finding the paper does not report, are also written into an automated check that runs before every release and flags that wording on any page. The core studies behind the measurements are listed in the research base, with the pages that use each one.
The one number we call validated, Your Face Score, has its test published in full — method, results and limits — on the Face Score validation page. Articles show the date they were last updated beside Randy's byline.
Contact
Questions about a result, a metric you don't understand, or feedback on the tools — we read every message and respond within 24 hours.
See what your face is actually doing
17 metrics, real measurements, no fluff. Takes about 60 seconds.
No signup required. Photos processed in memory and deleted instantly — nothing is kept unless you choose to save it.