Your photo never leaves your device
This face shape detector measures shape, proportions and eye angle from 478 points in your browser. Free, and the photo stays on this device.
Tick the box, then use a front-facing photo: even light, relaxed mouth
No photo needed. An AI-generated example, already measured.
Every measurement on this page is free. No account, no card.
AI-generated example portraitThree steps
The analysis runs in your browser. Your photo and your measurements stay on this device.
Use your camera or a front-facing photo in even light, hair off the forehead. It opens in your browser and is never sent anywhere.
Google's MediaPipe Face Landmarker runs on your device, corrects for head tilt and maps your features five times over.
Face shape, proportions, symmetry, eye angle and jaw angle, with the reasoning behind each number. Free, no account.
Measured examples
Each portrait below is AI-generated, not a real person. We ran the same on-device detector you use and show exactly what it returned. Tap a measurement to see where it is taken on the face.
Each measurement gets its own tool:Golden ratio face →Face symmetry test →Canthal tilt test →Nose shape →
AI-generatedMeasured shape
Jaw vs cheekbones = jaw width as a share of cheekbone width. Golden ratio gap = average distance from the classic 1.618 proportions.
Frames our rules suggest (a styling opinion, not a measurement): Aviator, Round, Oval
AI-generatedMeasured shape
Jaw vs cheekbones = jaw width as a share of cheekbone width. Golden ratio gap = average distance from the classic 1.618 proportions.
Frames our rules suggest (a styling opinion, not a measurement): Round, Oval, Cat-eye
AI-generatedMeasured shape
Jaw vs cheekbones = jaw width as a share of cheekbone width. Golden ratio gap = average distance from the classic 1.618 proportions.
Frames our rules suggest (a styling opinion, not a measurement): Round, Oval, Cat-eye
AI-generatedMeasured shape
Jaw vs cheekbones = jaw width as a share of cheekbone width. Golden ratio gap = average distance from the classic 1.618 proportions.
Frames our rules suggest (a styling opinion, not a measurement): Oversized, Deep square, Aviator
AI-generatedMeasured shape
Jaw vs cheekbones = jaw width as a share of cheekbone width. Golden ratio gap = average distance from the classic 1.618 proportions.
Frames our rules suggest (a styling opinion, not a measurement): Oval, Round, Aviator
AI-generatedMeasured shape
Jaw vs cheekbones = jaw width as a share of cheekbone width. Golden ratio gap = average distance from the classic 1.618 proportions.
Frames our rules suggest (a styling opinion, not a measurement): Square, Aviator, Browline
These are AI faces. Yours takes about two seconds.
Measure my faceFace shape guide
A face shape comes from four numbers: face length, and the widths of the forehead, cheekbones and jaw. Most people sit between two shapes, so treat the label as a starting point and look at the measurements behind it.
AI-generatedMeasured by this tool: 1.37 : 1 length to width, jaw angle 131°. See this face in the gallery →
Longer than it is wide, with the widest point at the cheekbones and a jaw that narrows gently into a rounded chin.
How to spot it: Proportions sit in the middle of most ranges, which is why oval is often called the easiest to style.
If this face wanted a new cut: Keep the crown flat and add width at the sides, with ends at or above the chin and soft layers from the jaw down.
Aviator, Round, Oval frames.
Suggestions come from the measurements, not the label, using the same rules as your result, so two people with the same shape can get different advice.
Not sure which one you are? The tool decides from your own measurements.
Measure my face| Shape | Length vs width | Widest point | Jaw |
|---|---|---|---|
| Oval | About 1.3 to 1.5× the cheekbone width once the hairline is included | Forehead a little wider than the jaw; cheekbones widest | Rounded, no strong corners |
| Round | Close to 1× the cheekbone width | Forehead, cheekbones and jaw fairly similar, cheeks fullest | Soft, wide jaw angle |
| Square | About 1 to 1.2× the cheekbone width | All three widths close together | Sharper jaw angle, straight jawline |
| Oblong | Around 1.5× the cheekbone width or more | Forehead, cheekbones and jaw similar; sides fairly straight | Can be soft or angular |
| Heart | About 1.3 to 1.5× the cheekbone width | Forehead widest, jaw clearly narrowest | Narrow, tapering to the chin |
| Diamond | About 1.3 to 1.5× the cheekbone width | Cheekbones widest; forehead and jaw both narrower | Narrow, often angular chin |
The question everybody asks
There is no peer-reviewed prevalence distribution for oval, round, square, heart, diamond, oblong and triangle in any human population. Not a small one, not an old one. Every percentage you have read comes from a face-shape tool's own user logs, and they contradict each other so violently that they cannot all be describing the same species.
| Site | Faces claimed | Oval | Diamond | Source cited |
|---|---|---|---|---|
| thefacereport.com | 15,429 | 49.2% | 5.4% | none |
| faceauraai.com (newer) | 18,041 | 38.8% | 25.0% | none |
| faceauraai.com (older) | 3,803 | ~46% | 22% | none |
| oblongfaceshape.com | 51,247+ | 28.4% | ~5% | Farkas 1994 |
Diamond ranges from about 5% to 25%, a five-fold disagreement, and one site calls it the rarest shape while another's own data puts it second most common. Triangle is reported at 0.03%, 0.05% and 8.6% by three different tools. The one citation anybody offers, Farkas 1994, is an anthropometry reference that publishes linear and angular norms; it does not contain frequencies of "oval" or "heart" at all.
Worth saying who agrees: one of those sites states plainly that face-shape categories are "a styling convention, not a rigorous scientific construct." That is correct, and it sits on the same site as a 49.2% figure.
Not from anatomy. The machine-vision literature on face shape uses five classes: heart, oblong, oval, round, square. Triangle and diamond are additions from styling sites, which is why their reported prevalence swings by two orders of magnitude. The taxonomy itself traces to eyewear retail: a published survey of the field notes that OPSM Opticians recognised four "traditional and archaic" shapes, then commissioned work in 2014 proposing five more (kite, rectangle, teardrop, heptagon and oblong), a nine-class scheme that it records as "yet to become mainstream." The number of face shapes is unsettled at four, five, six, seven or nine depending on who is selling what.
Physical anthropology has measured face proportions for a century, using the morphological facial index: face height divided by cheekbone width, times 100, with five named bands. It is replicable, caliper-measurable and has real population distributions. Here is what it finds.
Nepal, N=173
dental students 17 to 25, sliding caliper
Leptoprosopic
40.5%
most common band
India, N=200
medical students, mean age 20.1
Mesoprosopic
34.5%
most common band
Latvia, N=375
healthy children aged 1 to 15
Hypereuryprosopic
45.5%
most common band
Three samples, three different answers. The most common face is long and narrow in the Nepali students, middling in the Indian students, and short and wide in the Latvian children, where the index also climbs steadily with age, from 70.1 at age one to 82.3 by fifteen. There is no single most common face. There is a distribution that moves with population and with age, which is a real answer, and it is the opposite of a league table.
About half the time. A University of the Philippines study building a face-shape classifier checked celebrities whose shape is published in multiple places, and found that 62 of 117 names had conflicting classifications across the 21 sites reviewed. Same faces, same seven labels, 53% disagreement. That is the measured version of the thing competitors gesture at when they say "most people fall between two shapes" without ever putting a number on it.
The same paper is candid about where an "in-between" threshold comes from: a face is called a blend when a blending score lands between 40% and 60%. That is a number an engineer chose. And the best published classifier in that work agrees with its own human labellers only about 85% of the time, on five classes rather than seven. Any page handing you a seven-way categorical verdict as fact is claiming a precision no published method has demonstrated.
This is why the result above shows you every shape's similarity rather than one label, and why a borderline face is reported as borderline. The uncertainty is real, it is documented, and hiding it behind a single word would be the easiest thing on this page to get wrong.
Sources: Shrestha et al., JNMA 2019 (Nepal, N=173) · Choudhary et al., Cureus 2026 (India, N=200) · Grabcika et al., Children 2024 (Latvia, N=375) · Tio, "Face shape classification using Inception v3," University of the Philippines Diliman (N=500). Tool-site figures as published by each site on the date reviewed.
No app needed
All you need is a soft measuring tape, or a mirror and a washable marker to trace your outline. The tool above does the same comparison from 478 points, but measuring by hand is a good way to check a result that surprises you.
Tracing tip: stand still, close one eye, and draw around your face on the mirror from hairline to chin. Step back and compare the outline with the shape guide above.
AI-generatedPull all your hair back off your forehead and away from your jaw. Stand square to a mirror, or take a front photo at arm's length with the camera at eye level.
Measure across the widest part of your forehead, roughly halfway between your eyebrows and your hairline, from one side to the other.
Find the most prominent point of each cheekbone, just below the outer corner of each eye, and measure straight across between them.
Measure straight across between the two points below your ears where the jaw turns upward. That is the jaw width the tool compares with your cheekbones, not the length of the jawline itself.
Measure from the centre of your hairline straight down to the bottom of your chin. The tool can't see a hairline, so it adds an estimated 12% above the top of the face mesh to stand in for this.
Longest measurement is face length and it's about 1.5× the cheekbones? Oval or oblong. All widths similar with a sharp jaw? Square. Forehead widest? Heart. Cheekbones widest with a narrow forehead and chin? Diamond.
Beyond face shape
Eye shape names describe the opening, the corner angle and the eyelid. From a front photo we measure the first two.
Most nose type names describe the bridge, which is a profile feature. A front photo can still measure nose width in proportion to your eyes and face.
From a front photo we can measure nose width against your eye spacing and face width. Bridge shape (Greek, Roman, aquiline) needs a side photo, so we don't guess at it.
Method and accuracy
This is an estimate for styling and curiosity. It hasn't been validated against professionally labelled photos, and we'd rather say so than show a confident number that isn't earned.
Privacy
Face geometry is sensitive, so the tool is built to work without a server: the model downloads once from this site and runs in your browser tab.
Free tools
Every tool uses the same on-device measurements. The face shape detector is live now; the others open one at a time, and their pages explain what each will measure.
FAQ
Upload or take a front-facing photo with your hair pulled back and the tool measures your face length against the widths of your forehead, cheekbones and jaw. It shows the closest shape, the next closest, and the measurements behind both.
Pull your hair back and measure your forehead, cheekbones, jawline and face length with a soft tape, or trace your outline on a mirror. Then compare which measurement is largest. The step-by-step guide on this page walks through it.
Treat any face shape detector as an estimate, including this one. It has never been checked against professionally labelled photos, so we can't tell you how often the label is right. The hairline is an estimate, a smile or a beard moves the jaw points, and faces near a boundary are shown as two shapes rather than forced into one.
There's no reliable population study we can point to. Styling guides often call oval the most common, but that isn't based on measured data, and our 22-face test set isn't large or representative enough to say.
Your bone structure is largely set in adulthood, but weight changes, ageing and changes in fullness around the cheeks and jaw can change how your face measures. A beard or a hairstyle changes how the shape looks without changing the measurements underneath.
Take glasses off and pull bangs back: frames cover the eye corners and hair hides the forehead width. A full beard can shift the jaw points, so treat jaw angle and square or round results with extra caution.
Face shapes are a spectrum. When the two closest matches are near each other, we show both rather than force one label, and the styling tips come from your measurements instead of either label.
Canthal tilt is the angle of the line from the inner corner of the eye to the outer corner. Positive means the outer corner sits higher, negative means lower. It describes eye angle; it isn't a rating.
The golden ratio (about 1.618) is a classic proportion. The tool compares six face ratios with their classic values and reports the average gap. It describes proportion, not attractiveness, and a larger gap is common and normal.
No. The face model downloads to your browser and runs there. Your photo is read from memory and discarded when you close the tab. The planned paid report isn't available yet; when it launches it will be written from the numbers only.
Face geometry counts as biometric information in some places. We only process it on your device and with your consent, and the tool is meant for adults.
No. It is a styling and curiosity tool for haircuts, frames and makeup ideas, not a medical or cosmetic assessment.
Free, private and on your device. No upload, no sign-up.