HOW IT WORKS

The first rule of Beauty is the Golden Rule

We trace your face, measure a few simple proportions. That's it.

Drag across the photo to see the outline we trace
IN 30 SECONDS

Three easy steps.

1

Look at the camera

Sit in good light and look straight ahead. We'll ask you to turn a little and smile once.

2

We trace your face

Hundreds of tiny dots land on your eyes, nose, lips and jaw, like a connect-the-dots drawing of you.

3

See your results

You get your outline, seven measurements and your score out of 100.

WHAT WE LOOK AT

A few simple proportions.

Everything is compared to the size of your own face, so it doesn't matter how close you sit.

Pupil spacing

The distance between your pupils as a percentage of your face width.

Left–right balance

How closely your left and right sides mirror each other.

Eye gap

The gap between your eyes, compared with the width of one eye.

Mouth & nose

How wide your smile line is, next to how wide your nose is.

Lip balance

Your lower lip compared with your upper lip.

Eye tilt

Whether the outer corners of your eyes tilt up or down.

Face shape

The overall shape of your face: wider or longer.

WHAT SCIENCE SAYS

The big picture.

A face is more than its parts

People judge the whole face, including skin, expression and personality, not just measurements. [P4][P12]

There's no magic ratio

Faces near the average, and fairly symmetric ones, tend to be liked; the golden-ratio “perfect face” doesn’t hold up. So we never compare you to an ideal. [P11][P13][P2]

Photos are moments

Lighting, camera angle and a tilt of the head all change how a face looks and measures. [P7][P24]

YOUR PRIVACY

Your face stays with you.

Runs on your device

All the tracing happens in your browser. Nothing is uploaded.

No tracking

No analytics, no face recognition. Save the PDF with Google and we keep your email and report numbers, never photos.

Gone in 24 hours

Your report lives in this tab only, and you can clear it anytime.

Ready to rate your geometry?

COMMON QUESTIONS

Good to know.

Is HOT HOT PASS free?

Yes. The scan, the score, the colour study and the full web report are free. Saving your report as a PDF asks you to sign in with Google.

Does it upload my face?

No. Face tracing and every measurement run in your browser, and camera video is never stored or uploaded. If you sign in to save the PDF, we keep your name, email and your report’s numbers (score, measurements and colour readings) with your account. Photos and face outlines are never sent.

How is the score worked out?

A small published model of 24 decision trees reads five proportions (face width ÷ upper-face height, mouth width ÷ pupil spacing, face length ÷ width, inner-brow gap ÷ pupil spacing and temple width ÷ face width), removes the effect of head tilt and an open mouth, and maps the result so 75 is typical. On faces it never trained on, it agrees with two public rating guides only moderately (rank correlation about 0.5). Full details.

Does a high score mean I’m more attractive?

No. Studies find that faces close to the average and fairly symmetric tend to be liked, but no single ratio predicts beauty, and the famous golden-ratio “perfect face” does not hold up. Read the score as a playful estimate. [P11][P12][P13]

What is canthal tilt?

The angle between the inner and outer corners of each eye. HOT HOT PASS reports it as a 2D camera estimate, corrected for head roll. Head pitch changes how it looks, and raising the outer corners does not make faces more attractive in rating studies, so no ideal angle is given. [P24][P25][P26]

Which devices does it work on?

Any recent phone, tablet or computer with a camera and an up-to-date browser such as Safari, Chrome, Firefox or Edge. Good, even light gives the steadiest measurements.

FOR THE CURIOUSTechnical definitions & the research

Exact definitions

All measurements are approximate 2D image proportions, not millimetres. We don't measure skin health, tooth alignment, hairline ratios or demographics.

Pupil spacing / face width

Distance between pupil centres divided by visible face width (landmarks 234–454), expressed as a percentage. When iris landmarks are unavailable, the midpoint of each eye’s corners estimates its centre.

A camera proportion with different boundaries from Pallett’s inner-ear face width; 46% is not a target for this value. [P3][P18][P21]

Bilateral deviation

Mean reflected displacement of 8 eye, brow, nose, and mouth landmark pairs / face span, after roll correction.

Describes local landmark alignment. It excludes the jaw pairs, remains sensitive to pose, and has no validated attractiveness weight. [P2][P12][P23]

Inner-eye spacing

Inner-canthus gap / mean eye-corner width.

A descriptive proportion. The supplied papers do not validate 1:1 as an individual optimum. [P2][P3][P18]

Mouth / nose width

Mouth-corner span / alar-landmark span.

Expression-sensitive. Feature-rating correlations do not support a 1.55 target or a penalty for departing from it. [P1][P2][P18]

Lower / upper lip

Lower / upper vermilion landmark distance; not lip volume or total lip area.

Expression and landmark placement matter. The review reports conflicting lip preferences; no universal ratio is supported. [P1][P2][P18]

Canthal tilt

Mean outer-eye elevation relative to the inner corners, corrected for estimated face-axis roll.

A 2D angle proxy, affected by yaw, pitch, and corner localization. No ideal angle is established here. [P2][P7][P24][P25][P26]

Upper-face width / height

Landmark 234–454 span / vertical landmark 9–0 distance.

A mesh proxy, not clinical fWHR or the 36% hairline-based ratio. It does not establish attractiveness or personality. [P2][P3][P19][P20]

How the score is calculated

Skin colour and teeth are excluded from the score. The score is a small, published model that estimates how two public online rating guides (one for women, one for men) rank faces, from five proportions: face width ÷ upper-face height (fWHR), mouth width ÷ pupil spacing, face length ÷ width, the gap between the inner brows ÷ pupil spacing, and temple width ÷ face width. These were picked because they track both guides and differ little between women and men, so the score needs no guess about your sex. Each proportion is compared with the guides’ faces, the part explained by head tilt or an open mouth is removed, and 24 small decision trees average their estimates. That estimate is mapped to a 1–10 scale centred on 7.5 (shown ×10, so 75 is typical). On faces it never trained on, it agrees with the guides only moderately (rank correlation about 0.5 for each; roughly a quarter of their variation). The guides are one online community’s opinion and include photo and grooming effects, so read the score as a playful estimate, not a measure of worth. The golden ratio (1.618) on screen is decoration and doesn’t count. [P11][P12][P13]

Research papers

The papers behind HOT HOT PASS, with what each one does and doesn't show. P1–P10 shaped the measurements; P11–P26 add the evidence for averageness, the golden-ratio critique, the care tips in your report, and measurement methods.

  1. P1

    Correlation between facial attractiveness and facial components assessed by laypersons and orthodontists ↗

    Ren, H.; Chen, X.; Zhang, Y. (2021)

    Journal of Dental Sciences, 16, 431–436

    Ratings of individual facial regions related to overall ratings, with different relationships for laypeople and orthodontists.

    Interpretation limit. Subjective feature ratings are not anatomical measurements or weights for a scoring formula.

    Association
  2. P2

    Art or Science? An Evidence-Based Approach to Human Facial Beauty a Quantitative Analysis Towards an Informed Clinical Aesthetic Practice ↗

    Harrar, H.; Myers, S.; Ghanem, A. M. (2018)

    Aesthetic Plastic Surgery, 42, 137–146

    Quantitative facial preferences vary across studies; symmetry, lip proportions, and classical ratio findings are inconsistent.

    Interpretation limit. A heterogeneous review does not establish universal targets or validate these camera proxies.

    Review
  3. P3

    New “Golden” Ratios for Facial Beauty ↗

    Pallett, P. M.; Link, S.; Lee, K. (2010)

    Vision Research, 50(2), 149–154

    Manipulated photographs of White female faces showed preferred eye-to-mouth/face-length and pupil-spacing/face-width regions near 36–37% and 46%.

    Interpretation limit. The study used hairline-to-chin and inner-ear boundaries. HOT HOT PASS does not measure those boundaries; its ratios cannot be compared with these targets.

    Image experiment
  4. P4

    The Relative Contributions of Facial Shape and Surface Information to Perceptions of Attractiveness and Dominance ↗

    Torrance, J. S.; Wincenciak, J.; Hahn, A. C.; DeBruine, L. M.; Jones, B. C. (2014)

    PLOS ONE, 9(10), e104415

    Both shape and surface information contributed to attractiveness ratings in the study images.

    Interpretation limit. Surface includes texture, colour, contrast, and shading. The findings do not provide weights for a geometry-only app.

    Image decomposition / ratings
  5. P5

    Your face looks the same as before, only prettier: The facial skin homogeneity effects on face change detection and facial attractiveness perception ↗

    Sun, Y.-H. P.; Zhang, X.; Lu, N.; Li, J.; Wang, Z. (2022)

    Frontiers in Psychology, 13, 935347

    Digitally increasing skin texture/colour homogeneity increased attractiveness judgments in the experimental tasks.

    Interpretation limit. The experiment did not test skincare products or isolate colour from texture. Four median swatches do not measure homogeneity.

    Image experiment
  6. P6

    Understanding Beauty via Deep Facial Features ↗

    Liu, X.; Li, T.; Peng, H.; Ouyang, I. C.; Kim, T.; Wang, R. (2019)

    arXiv:1902.05380v2

    Classifier attributes were associated with image ratings; a separate preference task evaluated generated image edits.

    Interpretation limit. Dataset bias, correlated attributes, and synthetic edits prevent universal or causal styling prescriptions.

    Computational / preprint
  7. P7

    Human Face Tilt Is a Dynamic Social Signal That Affects Perceptions of Dimorphism, Attractiveness, and Dominance ↗

    Marshall, P.; Bartolacci, A.; Burke, D. (2020)

    Evolutionary Psychology, 18(1), 1–15

    Changing head pitch and gaze affected judgments of rendered faces.

    Interpretation limit. Synthetic stimuli and context-specific effects do not establish one flattering angle for everyone.

    Rendered-face experiment
  8. P8

    The Role of the Posed Smile in Overall Facial Esthetics ↗

    Havens, D. C.; McNamara, J. A., Jr.; Sigler, L. M.; Baccetti, T. (2010)

    The Angle Orthodontist, 80(2), 322–328

    Pre/post orthodontic photographs showed stronger evidence for improved smile-region ratings than whole-face ratings.

    Interpretation limit. There was no untreated comparison. The work does not show that everyone should alter their smile or whiten their teeth.

    Retrospective association
  9. P9

    The Influence of Anteroposterior Head Inclination on the Perceived Consonance of the Smile Arc and Lower Lip Curvature on Photographs: A Cross-Sectional Study ↗

    Khadem, A. R.; Togninalli, M.; Antonarakis, G. S.; Vento, C. (2025)

    Journal of Clinical Medicine, 14, 1658

    Head inclination changed the measured relationship between the photographed smile arc and lower-lip curvature.

    Interpretation limit. There were no attractiveness raters. HOT HOT PASS has no tooth-edge landmarks and does not measure smile-arc consonance.

    Photographic geometry
  10. P10

    Tooth Colour and Facial Attractiveness: Study Protocol for Self-Perception with a Gender-Based Approach ↗

    Mazur, M.; Jedliński, M.; Westland, S.; Piroli, M.; Luperini, M.; Ndokaj, A.; Janiszewska-Olszowska, J.; Nardi, G. M. (2024)

    Journal of Personalized Medicine, 14, 374

    The paper proposes studying perceived tooth colour and instrumental colour measurements.

    Interpretation limit. It contains no outcome data and cannot establish a whitening benefit, shade target, or validated camera score.

    Protocol / no results
  11. P11

    Attractive Faces Are Only Average ↗

    Langlois, J. H.; Roggman, L. A. (1990)

    Psychological Science, 1(2), 115–121

    Digitally averaged composite faces were rated more attractive than most of the individual faces that made them, and more so as more faces were averaged.

    Interpretation limit. Averaging also smooths skin and removes blemishes. Being close to a population average is one factor, not a target to measure yourself against.

    Image experiment · strong
  12. P12

    The Evolutionary Psychology of Facial Beauty ↗

    Rhodes, G. (2006)

    Annual Review of Psychology, 57, 199–226

    Averageness, symmetry and sex-typical traits are consistently, if modestly, preferred across cultures; their link to actual health is weaker than often claimed.

    Interpretation limit. Preferences are statistical tendencies across raters, not rules for an individual face.

    Review · strong
  13. P13

    Marquardt’s Phi Mask: Pitfalls of Relying on Fashion Models and the Golden Ratio to Describe a Beautiful Face ↗

    Holland, E. (2008)

    Aesthetic Plastic Surgery, 32(2), 200–208

    Procrustes analysis shows the golden-ratio “Phi mask” is assessed with a faulty fit method, suits non-European faces poorly and resembles a masculinised female face.

    Interpretation limit. A critique of one template; it does not supply a better ideal. It is why HOT HOT PASS scores no golden ratio.

    Shape-analysis critique
  14. P14

    Sunscreen and Prevention of Skin Aging: A Randomized Trial ↗

    Hughes, M. C. B.; Williams, G. M.; Baker, P.; Green, A. C. (2013)

    Annals of Internal Medicine, 158(11), 781–790

    In 903 adults, daily broad-spectrum sunscreen led to measurably less skin ageing over 4.5 years than discretionary use.

    Interpretation limit. Skin ageing was graded from skin-surface casts, not attractiveness ratings; participants were adults under 55 in sunny Australia.

    Randomised trial · strong
  15. P15

    Comparative efficacy of topical interventions for facial photoaging: a network meta-analysis ↗

    Lin, L.; Chen, X.; Liu, C.; Wang, Q.; Lian, W.; Xu, X.; Guo, Y.; et al. (2025)

    Scientific Reports, 15, 26889

    Across 23 randomised trials (3,905 people), retinoids such as tretinoin, retinol and isotretinoin improved fine wrinkles, tazarotene coarse wrinkles, with differing tolerability.

    Interpretation limit. Limited racial diversity and possible commercial bias. Prescription retinoids need a clinician; none is suitable in pregnancy.

    Network meta-analysis · strong
  16. P16

    Color homogeneity and visual perception of age, health, and attractiveness of female facial skin ↗

    Matts, P. J.; Fink, B.; Grammer, K.; Burquest, M. (2007)

    Journal of the American Academy of Dermatology, 57(6), 977–984

    Skin with more even colour (less uneven pigmentation and redness) was judged younger, healthier and more attractive, independent of face shape.

    Interpretation limit. Rated women’s skin only, with images in which colour was altered digitally; the app’s four median swatches do not measure evenness.

    Image experiment · moderate
  17. P17

    You Are What You Eat: Within-Subject Increases in Fruit and Vegetable Consumption Confer Beneficial Skin-Color Changes ↗

    Whitehead, R. D.; Re, D.; Xiao, D.; Ozakinci, G.; Perrett, D. I. (2012)

    PLOS ONE, 7(3), e32988

    Eating more fruit and vegetables over six weeks made skin measurably more yellow/golden (carotenoids), a change raters see as healthier.

    Interpretation limit. A small, short, mostly young White sample; not a diet plan and not tested against a control group.

    Within-person diet study · moderate
  18. P18

    Anthropometry of the Head and Face (2nd ed.) ↗

    Farkas, L. G. (ed.) (1994)

    Raven Press, New York

    Standard definitions of facial landmarks, distances and indices used in clinical face measurement.

    Interpretation limit. Defined for calipers on a person in a set head position. Camera landmarks only approximate them, so values are not interchangeable.

    Reference text
  19. P19

    Evidence from Meta-Analyses of the Facial Width-to-Height Ratio as an Evolved Cue of Threat ↗

    Geniole, S. N.; Denson, T. F.; Dixson, B. J.; Carré, J. M.; McCormick, C. M. (2015)

    PLOS ONE, 10(7), e0132726

    fWHR relates weakly to perceived and actual aggression, with small, variable effects across studies.

    Interpretation limit. Effects are small and context-dependent; fWHR is not an attractiveness measure, which is why it is only one input to the score.

    Meta-analysis
  20. P20

    In your face: facial metrics predict aggressive behaviour in the laboratory and in varsity and professional hockey players ↗

    Carré, J. M.; McCormick, C. M. (2008)

    Proceedings of the Royal Society B, 275(1651), 2651–2656

    Early study defining photo-based fWHR (bizygomatic width ÷ brow-to-upper-lip height) and linking it to aggression in men.

    Interpretation limit. Small samples and later mixed replications. Relevant for the measurement definition, not as an attractiveness model.

    Association · historical
  21. P21

    Variation and extrema of human interpupillary distance ↗

    Dodgson, N. A. (2004)

    Proceedings of SPIE 5291, Stereoscopic Displays and Virtual Reality Systems XI, 36–46

    Adult pupil distance averages about 63 mm with most people between roughly 50 and 75 mm.

    Interpretation limit. Physical millimetres; the app measures pupil spacing as a share of face width and uses it to scale other measures.

    Reference data
  22. P22

    MediaPipe: A Framework for Building Perception Pipelines ↗

    Lugaresi, C.; Tang, J.; Nash, H.; et al. (2019)

    arXiv:1906.08172

    Describes the on-device pipeline framework behind the Face Landmarker that traces the 478 points HOT HOT PASS uses.

    Interpretation limit. Landmarks are model estimates; accuracy varies with light, pose, occlusion and face shape.

    Software framework
  23. P23

    Facial asymmetry in young healthy subjects evaluated by statistical shape analysis ↗

    Ercan, I.; Ozdemir, S. T.; Etoz, A.; Sigirli, D.; Tubbs, R. S.; Loukas, M.; Guney, I. (2008)

    Journal of Anatomy, 213(6), 663–669

    In 321 healthy young adults, measurable left–right asymmetry was normal, with the left side most often dominant.

    Interpretation limit. Supports measuring asymmetry objectively and treating some asymmetry as normal; it gives no attractiveness weight.

    Shape analysis
  24. P24

    The Influence of Sagittal Head Tilt on Periorbital Appearance: Implications for Clinical Photography and the Evaluation of Postoperative Results ↗

    Vaca, E. E.; Bricker, J. T.; Mioton, L. M.; Fagien, S.; Alghoul, M. S. (2021)

    Aesthetic Surgery Journal Open Forum, 4, ojab043

    Tilting the head down (−15° to +15° tested) changed measured eye-area features, made canthal tilt more positive and was rated more attractive.

    Interpretation limit. 12 women, eye crops only. Shows that pose changes eye measurements, so the app corrects for head pitch.

    Photo experiment
  25. P25

    Diagnostic analysis of vertical orbital dystopia and canthal tilt for surgical correction ↗

    Lee, J.-Y.; Choung, H.-W.; Choung, P.-H. (2020)

    Journal of the Korean Association of Oral and Maxillofacial Surgeons, 46(6), 379–384

    Measured canthal tilt as the inner-to-outer corner angle: about 8.4° on average in 76 Korean adults, with small left–right eye-level differences.

    Interpretation limit. One population and clinical photographs; a reference for the measurement, not an ideal angle.

    Clinical measurement
  26. P26

    Understanding the impact of eyebrow and lateral canthal height on the perception of character traits: A rating-based analysis of 2081 participants ↗

    Kempa, J.; Kasielska-Trojan, A.; Gabryszewski, M.; Antoszewski, B.; Biesman, B. S.; Knoedler, L.; Cotofana, S.; Alfertshofer, M. (2025)

    Journal of Plastic, Reconstructive & Aesthetic Surgery, 104, 417–425

    Digitally raising the eyebrows or outer eye corners lowered attractiveness, femininity and trustworthiness ratings and raised perceived dominance.

    Interpretation limit. Edited images of a few faces; supports “no higher is better” for canthal tilt, not a target angle.

    Rating experiment

Landing sculpture: Head Of Michelangelo's David, Optimised by Thomas Flynn / CC BY 4.0. Material, lighting, outline overlays, and scan animation adapted for HOT HOT PASS. Tracking: MediaPipe Face Landmarker [P22].

HOT HOT PASS is a free face-geometry study that runs entirely in your browser. It traces 478 facial landmarks, measures seven proportions (pupil spacing, left–right balance, eye gap, mouth-to-nose width, lip balance, canthal tilt and face width-to-height) and explains each one with published research. Camera video never leaves your device.

Last updated · Method version 2.1