About us
A styling tool that shows its work.
Real Face Shape Detector helps you explore an estimated facial shape and turn it into practical ideas for hairstyles, eyewear, grooming, colour, and personal styling. The analysis runs in your browser, and the report explains the measurements and uncertainty behind the result.
Why this tool exists
Face-shape advice is often either too vague to use or too confident to trust. A useful middle ground is a transparent estimate: show the visible proportions, name the closest category and a plausible alternative, flag problems with the photo, and explain why a styling suggestion might work.
That is the purpose of this site. It is a free, browser-based styling reference for people who want a clearer starting point before choosing a haircut, pair of glasses, beard outline, neckline, or colour direction. It is not a test you pass or fail, and the result is not a verdict about your appearance.
The site also takes a deliberately local approach to a sensitive kind of image. The detector does not require an account, and the application code is designed to analyse the photo on your device rather than upload it. The How it works page describes the pipeline in detail, and the Privacy Policy explains the storage and email practices that remain.
How the tool works, at a high level
- You choose a photo or capture one with the optional live camera. The browser decodes the image locally.
- A bundled MediaPipe Face Landmarker model finds facial landmark points. The model and runtime are served from this website, not requested from an external AI service.
- The app aligns the eye line, calculates normalized widths, lengths, taper, and other visible ratios, and checks capture quality.
- Those measurements are compared with seven face-shape prototypes. The report shows the closest reading, alternatives, relative geometry fit, and any quality warnings.
- Fixed styling rules turn the estimate and optional preferences into hair, frame, grooming, colour, and outfit suggestions. You can override the shape reading.
There is no trained face-shape image classifier behind the label and no generative recommendation model. The limits of the geometric approach, including the lack of an expert-labelled accuracy benchmark, are documented openly on the methodology page.
The short version
Measurements, not a beauty score.
The report estimates visible proportions. It does not rate attractiveness, diagnose a condition, determine identity, or claim to know your bone structure from a single photograph.
The category set
Seven shapes, treated as a spectrum.
The implementation defines seven categories. They are styling labels, not scientific classifications, and many faces sit between two of them. Each guide explains the visible pattern and common alternatives.
Oval
Balanced and adaptable
Open the oval guide →
Round
Soft curves, compact proportions
Open the round guide →
Square
Strong jaw, clear angles
Open the square guide →
Heart
Wider upper face, tapered chin
Open the heart guide →
Diamond
Prominent cheekbones, narrow ends
Open the diamond guide →
Oblong
Longer lines, straighter sides
Open the oblong guide →
Triangle
Strong lower face
Open the triangle guide →
Compare all seven
See the width, length, and jaw relationships side by side.
Open the shape library →
What the estimate cannot tell you
A face shape is an exploratory summary of visible 2D proportions. It is not a stable identity, a medical finding, or a precise anthropometric measurement. Facial shapes overlap, and a small change in head angle or image quality can move the closest reading.
- Camera angle, lens distortion, hairstyle, glasses, facial hair, makeup, and cropping can change the apparent proportions.
- The result depends on image quality and the landmark model’s confidence. Always check the capture-quality warnings.
- No accuracy benchmark against expert labels is claimed, and there is no representative demographic validation.
- A styling suggestion is a starting point, not a rule. Comfort, hair texture, frame fit, culture, and personal preference matter more than a category.
- The tool does not assess attractiveness, age, gender, ethnicity, health, or identity.
These are not fine-print caveats added afterward; they are part of the product design. The report is meant to help you ask better questions, not to end the conversation.
Privacy-first, within verified limits
The detector reads the selected file in the browser, processes the pixels locally with a bundled model, and keeps the photo in temporary memory. The application has no image upload endpoint and does not call an external AI provider. The photo is not written into a saved report or JSON export.
Some data does remain on the device: a theme preference, a session report for the current tab, and a saved report if you explicitly choose “Save report”. Email you send us is handled by the email providers involved. Hosting and security providers may process ordinary request metadata. The Privacy Policy describes each of these accurately rather than claiming that no data is ever processed.
There is no account system, no analytics script, and no advertising script in the current build. If advertising is added later, the privacy disclosures and any required consent mechanism will be updated before it goes live.
Where to go next
Start with the estimate, then use the guides.
The report is most useful when you compare it with the reasoning and try a few directions rather than treating the first label as fixed.
Hairstyle guide
Cuts, fringes, and volume directions by estimated shape.
Open the guide →
Glasses and sunglasses
Frame shapes, fit checks, and contrast notes.
Open the guide →
Beard and grooming
Length, outline, neckline, and maintenance directions.
Open the guide →
Outfit builder
A deterministic way to combine necklines, layers, and proportions.
Open the guide →
Clothing colours
Palettes keyed to a visible undertone estimate.
Open the guide →
Skin tone and colour
How the photo-based colour estimate works and where it stops.
Open the guide →
Questions and answers
Privacy, accuracy, photos, devices, and styling in one place.
Read the FAQ →
Corrections and feedback
Report a measurement issue, accessibility problem, or unclear claim.
Contact us →