There are six main eye shapes — almond, round, monolid, hooded, upturned, downturned — and almost everyone is a combination, not a pure example. You figure out yours with three quick checks in a mirror: is the crease visible, where does the outer corner sit relative to the inner, and how much white shows above and below the iris.
The longer version is more useful — the six categories layer with secondary descriptors (close-set, deep-set), the "most attractive" question has a more honest answer than Pinterest suggests, and AI is good at this read because it measures the exact landmarks oculoplastic surgeons use.
Eye shape comes out of oculoplastic anatomy. Three measurable features define the categories: the palpebral fissure (gap between upper and lower lids), the canthal tilt (the angle of a line drawn from inner to outer corner), and the supratarsal crease (the fold above the upper lash line, when one exists). Anthropometric studies measure all three directly — length, height, and inclination in degrees. The styling categories you see online are simplified shorthand for those measurements.
Visible iris at top and bottom, slight upward tilt at the outer corner, defined supratarsal crease. The canthal tilt is positive, typically between 0° and 5° — class I canthal tilt is the modal category in clinical classifications. Almond is also the shape most beauty editorial defaults to as "ideal," which is its own bias (covered below).
Whites visible above and below the iris, not just at the corners. Crease present, palpebral fissure taller relative to length, eye reads more circular than oval. Round eyes read younger partly because they share the neonate facial proportions Cunningham linked to perceived attractiveness in his multiple fitness model.
No visible supratarsal crease above the lash line. The lid runs from lash line to brow bone in a single plane, often with an epicanthic fold covering the inner corner. Prevalence varies sharply by population. A 3D photogrammetry study found all Malays and 70.1% of Chinese in Malaysia had a double eyelid, putting monolid prevalence in the Chinese sample around 30%, with mean crease height of 4.91 mm (vs 8.33 mm for Malays). The epicanthic fold itself is present in roughly 40–90% of East Asian populations and only 2–5% of non-Asian populations.
Excess skin from the brow bone covers part of the upper lid, so with the eye open and relaxed, less lid is visible than the lid actually contains. Hooded can be genetic (present from your teens) or acquired through aging — age-related hooding is called dermatochalasis and has a reported prevalence up to 17.8% in adults, increasing each decade with a heritability around 61% in twin studies. Most people who think their eyes "changed shape" in their 30s and 40s are seeing early dermatochalasis.
The outer corner sits noticeably higher than the inner corner. Canthal tilt is on the higher end, typically above 5°. Caucasian males show significantly larger canthal tilt than Chinese males in anthropometric comparisons, but upturned shape exists in every population — it is a degree on a continuum, not a binary.
The outer corner sits lower than the inner corner — negative canthal tilt, classified as class III in the surgical canthal tilt scale. Often confused with hooded eyes because both can make the outer eye area look "droopier," but they are anatomically different. Downturned is about corner position; hooded is about lid skin.
A note before you keep reading: most people are not pure examples of one shape. The honest read for many is "almond with mild hooding on the left," or "round with a slightly downturned outer corner." That is not indecision — it is just how anatomy works.
You need a mirror, decent lighting (front-facing, ideally a window), and your hair off your face. Relax your expression — no smile, no eyebrow lift.
Step 1: Crease check. Look straight ahead. Is there a visible fold of skin above your lash line where the upper lid meets the eye socket? If no crease at all → monolid. If a crease is present but the brow-bone skin droops over part of it → hooded. If the crease is present and clearly visible → continue.
Step 2: Corner angle. Picture an imaginary horizontal line through the center of your pupil. Where does your outer eye corner sit relative to your inner corner?
Step 3: White vs iris ratio. Look straight ahead and relax. Can you see white above AND below the iris, or just at the inner and outer edges?
That gets you a primary shape in under a minute. If you cannot tell between two — say, almond and round, or almond and slightly upturned — you are between them. Pick the closer one for styling purposes; both rules of thumb will mostly apply.
A lot of people search "how to determine eye shape from photo" because mirror angles are hard to read. Shortcut: take a front-facing photo at arm's length, eyes relaxed and looking straight at the lens, no smile, no eyebrow raise, diffuse natural light. Phone front cameras are wide-angle and distort up close — closer than ~30 cm changes apparent canthal tilt. Arm's length is the fair test.
The six categories cover the lid and corner. Three more descriptors layer on top.
Close-set vs wide-set. The classical anthropometric rule: the gap between your inner eye corners should equal the width of one eye. Less → close-set, more → wide-set. Studies measure this as intercanthal distance divided by palpebral fissure length, and variation is significant across populations.
Deep-set vs protruding. How far back the eyeball sits relative to the brow bone. Deep-set eyes appear shadowed; protruding eyes sit closer to the brow's vertical plane. Mostly a function of orbital bone structure and fat volume.
Asymmetry. Almost no one has identical eyes. Different crease heights, slightly different canthal tilts, one more hooded than the other — all normal. AI analyses of facial asymmetry routinely find measurable left/right differences in healthy faces and only flag them as clinically meaningful past a threshold.
The fully honest description for most people sounds like "almond + slightly close-set + mildly hooded on the left." That layered read is more useful than any single label.
Honest answer: there is no universal "rarest" because eye shape distributes very differently across populations. The monolid is one of the most common shapes in East and Southeast Asian populations and rare in European samples. The answer to "what's rarest" depends entirely on which population you ask about.
The clinical answer is that class III canthal tilt — genuinely downturned outer corners sitting below the inner corner — is the least common category in most anthropometric samples. But the data is thin and the "rarest" framing is mostly clickbait. Treat eye shape as a distribution, not a leaderboard.
There is no objectively most attractive eye shape. What the research actually shows:
What no study supports is a universal ranking like "almond > round > monolid." Western media has over-coded almond + symmetric + slight upturn as "ideal," but that is a cultural pattern, not a biological one. Different eras and cultures foreground different shapes — monolid in modern East Asian editorial, hooded in 1960s mod looks, round in the early-2010s dewy aesthetic. The defensible read: your face has shapes that flatter it more than others, but those depend on your other features, not on a global ranking.
AI eye shape detection runs the same measurements an oculoplastic surgeon would, automatically. First, facial landmark detection identifies anchor points — the standard dlib 68-point model places six points per eye, and MediaPipe's facemesh extends to 478 for finer arcs. From those points the model computes palpebral fissure length and height, canthal tilt, the eye aspect ratio (EAR), and presence/absence of the supratarsal crease and epicanthic fold.
Those measurements map cleanly to the six shape categories. Automated systems have been validated against expert oculoplastic measurements with reasonable agreement.
Where AI fails: closed or squinted eyes, heavy eye makeup (winged liner especially distorts apparent canthal tilt), harsh side lighting that creates fake shadow creases, low-resolution images (under ~60×60 px per eye drops accuracy noticeably), and anything that obscures the brow bone for hooding detection.
Want to know your eye shape without staring in a mirror? Our Eye Shape Analyzer runs the full landmark pipeline on a selfie and returns a primary shape, secondary descriptors (close-set, deep-set, asymmetry), and makeup suggestions. Free, no signup, instant. Pair it with the Eye Color Analyzer, Eyebrow Shape Analyzer, or Makeup Style Finder if you want more layers.
The geometry principle most makeup artists use: emphasize where the eye is already strong, or correct where the proportions feel off-balance — never both at once. Here is the practical version.
A lash map is the curl + length pattern across an eye, written inner → outer. Standard curls go J (least) → B → C → CC → D → DD (most lifted); lengths typically range 8–15 mm.
Same contrast principle as face shape: angular frames balance round eyes, softer frames balance angular features. Frame depth matters more than width for eye shape — shallow frames cut across the eye and shorten it; taller frames give it room.
Briefly: temporarily and subtly, yes — but not the way the question implies. Botox doesn't change your underlying eye shape. It shifts the position of soft tissue around the eye, which changes the apparent shape for the treatment window.
Two relevant mechanisms. Crow's feet botox targets the lateral orbicularis oculi, the muscle that closes the lid and pulls lateral skin into smile lines — relaxing it smooths the side of the eye and slightly opens the lateral palpebral fissure during smiling. A lateral brow lift technique (often combined with crow's feet treatment) produces 1–3 mm of lateral brow elevation, which de-hoods the upper lid temporarily.
A study tracking post-botox facial assessment found patients looked younger but not more attractive after lateral orbicularis treatment. The "younger" effect is the shape change people notice. The "not more attractive" finding is the reason to be calibrated — real but smaller than aesthetic-medicine marketing implies.
"Your eye shape is genetic and fixed." Partially true. The bone structure that determines orbit position is fixed in adulthood. Lid skin, brow position, and orbital fat all change — which is why hooding develops in nearly everyone past their 40s and 50s.
"Monolid means small eyes." False. Monolid refers to the absence of a supratarsal crease, not eye size. Monolid and double-lid populations show overlapping palpebral fissure length distributions in anthropometric studies.
"Eye yoga can change your eye shape." No credible evidence for permanent shape change. A small pre-experimental study found face yoga shifted measured orbicularis tonus and elasticity slightly, but without controls and without showing changed apparent eye shape. Treat "exercises for bigger eyes" the way you treat mewing — temporarily affects soft tissue tone, doesn't restructure anatomy.
"Eye shape predicts personality." That is physiognomy. Modern studies on face-shape-personality links don't survive replication, and the field has been debunked as pseudoscience since the mid-20th century. Treat any "what your eye shape says about you" content as a vibe quiz, not a finding.
Want the AI read on your own eyes? Start with the Eye Shape Analyzer and follow with the Face Shape Analyzer, Eye Color Analyzer, Eyebrow Shape Analyzer, and Makeup Style Finder. All free, no signup.