Skip to main content

Expression detail

Smile Detector and Smile Score

Updated 3 min read

The Smile analysis looks closely at one expression. It combines the happy score from the expression model with measurements from facial landmarks to describe a smile in parts: how strong it is, how much the mouth corners curve, how much of the teeth show, and how much the eyes narrow. That last one matters because a smile that reaches the eyes tends to read as more genuine.

It is a fun way to pick your best photo for a profile or ID-style picture, and to see why some smiles look forced. Everything runs in your browser through face-api.js on TensorFlow.js. The photo is never uploaded, and recent scans keep only the numbers, never the image.

Want every option? Open the Smile analysis

How to use it

  1. 1

    Take or choose a front-facing photo where the whole mouth and both eyes are visible and evenly lit.

  2. 2

    Add it to the Smile analysis and read the overall smile score, the happiness percentage and the natural smile score.

  3. 3

    Check the breakdown bars for mouth curve, teeth visibility and eye squint to see what drove the score.

  4. 4

    Try 2 or 3 shots with different smiles and keep the one whose eye squint and mouth curve look balanced.

What makes up the smile score

The happiness figure comes from the same expression network the Emotion analysis uses: it is how strongly the face matches the happy expression it learned from labelled photos. The other figures come from landmark geometry. Mouth curve measures how far the corners of the mouth lift relative to the centre of the lips. Teeth visibility looks at how far the lips part. Eye squint compares how open the eyes are, because the muscles of a full smile push the cheeks up and narrow the eyes.

The natural smile score leans on that last signal. Smiles that lift the mouth without narrowing the eyes are often described as polite or posed smiles, while smiles that crinkle the eyes are often described as genuine. The tool reports this as a pattern in the photo, not as a judgement of how anyone feels.

Getting a clear smile reading

  • Keep the head level: tilting it changes where the mouth corners sit relative to the lips.
  • Avoid hands, hair or a phone near the mouth.
  • Use soft front light so shadows do not hide the mouth corners.
  • A closed-lip smile scores low on teeth visibility by design; that is not a flaw.
  • Sunglasses hide the eyes, so the natural smile score is unreliable with them on.

Using it to pick your best photo

The most practical use is choosing between several shots taken a few seconds apart. Run each through the analysis and look at the breakdown rather than the headline. A photo with a high mouth curve but almost no eye squint often looks posed in a profile picture; one with balanced curve and squint usually reads as relaxed. If two shots score about the same, the scores have done their job and the choice is yours.

Photographers have long asked people to think of something funny instead of saying a word, because a laugh moves the cheeks and eyes as well as the mouth. You can try the same trick here: take a few frames with the camera, compare the natural smile scores, and see whether the advice holds for your face.

Limits

The scores describe one frame. A smile captured mid-movement can score lower than the same smile a fraction of a second later. People with facial differences, braces, facial paralysis or simply a different resting face can get readings that say nothing about their happiness. As with every analysis on this tool, use it for fun and for comparing your own photos, not for assessing other people.

Frequently asked questions

How is the smile score calculated?

It combines the happy expression score from a face-api.js expression model with landmark measurements of mouth curve, lip opening and eye narrowing.

What does the natural smile score mean?

It rises when the eyes narrow along with the mouth, a pattern often seen in spontaneous smiles. It is a description of the photo, not proof of how someone feels.

Why is my teeth visibility low?

Teeth visibility only measures how far the lips part. A closed-mouth smile scores low there and can still score well overall.

Can I use it to choose a profile photo?

Yes. Run two or three candidate photos and compare the breakdowns. Pick the one you like, too; the scores are a guide, not a rule.

Is my photo uploaded?

No. The analysis runs in your browser and the image is never sent anywhere.

Run a website? Link to this guide if it helps your readers, and see our free backlink sites list for places to get your own site listed.