Free tool · runs in your browser
The oldest rule of facial proportion: forehead, midface and lower face should each take about a third. Symmi measures your three as percentages and shows which one is long or short.
Drop a photo, paste one, or choose a file
Straight-on, whole face visible, even light. JPG, PNG or WebP.
Runs in your browser. Your photo never leaves your device.
How it is measured
The face is split at the brow line and the base of the nose. The hairline is estimated from the forehead landmarks (the mesh stops just below it) because hair covers it in most photos, so the upper third is the least certain of the three; the middle and lower thirds are measured directly from landmarks and are the readings the Symmi app scores.
Measure it yourself
Skip the ruler: the free tool above does it from one photo in about 30 seconds ↑
What's normal
Equal thirds would be 33% each. Faces rated attractive usually sit within a few points of that, with a slightly longer lower third on men and a slightly shorter one on women.
Sources
The measurement and its reference ranges come from published research. Studies use different methods and populations, so compare your number with the range shown on this page rather than with other tools.
Farkas LG, Hreczko TA, Kolar JC, Munro IR (1985). Vertical and horizontal proportions of the face in young adult North American Caucasians: revision of neoclassical canons. Plastic and Reconstructive Surgery. PubMed 3883374
Measured the vertical thirds on real faces; equal thirds were the exception, not the rule.
Farkas LG, Katic MJ, Forrest CR, et al. (2005). International anthropometric study of facial morphology in various ethnic groups/races. Journal of Craniofacial Surgery. PubMed 16077306
Facial proportions measured across ethnic groups; normal ranges vary by population.
Al-Sebaei MO (2015). The validity of three neo-classical facial canons in young adults originating from the Arabian Peninsula. Head & Face Medicine. PubMed 25889948
Tested the neoclassical canons in another population and found most did not apply.
Kartynnik Y, Ablavatski A, Grishchenko I, Grundmann M (2019). Real-time facial surface geometry from monocular video on mobile GPUs. arXiv. arXiv:1907.06724
The face-mesh model that places the 468 surface landmarks this tool measures from.
Ablavatski A, Vakunov A, Grishchenko I, et al. (2020). Real-time pupil tracking from monocular video for digital puppetry. arXiv. arXiv:2006.11341
The iris model that adds 10 eye landmarks (478 in total), used for pupil-based reference lines.
Our own data: how much each reading moves when the same face is photographed again is published on the methodology page. How we choose sources.
Related tools
FAQ
No. Faces rated attractive usually sit within a few points of equal, with a slightly longer lower third on men and a slightly shorter one on women. Ten points off on one third is what viewers notice.
Possibly not. The hairline is estimated, so treat the upper third as a rough number. The middle and lower thirds are measured and reliable.
Its bone length cannot. What can shift the reading is chin and jaw soft tissue and posture, both of which the Symmi routine targets. The app measures the lower third and the midface ratio on every scan so you can see whether it moved.
No. Everything runs in your browser.
A controlled scan, every reading against its ideal band, bone vs trainable, a plan, and a 3D model on Face ID iPhones.