Omogle face rating

Choose one front-facing photograph and the browser calculates a PSL-style score from six visible geometric measurements. The free result includes the score and three measurements; paid options explain all six, with a monthly toolkit for repeat scans.

The score is deterministic and the inputs are visible. It is still a playful heuristic rather than a scientific attractiveness test.

1. 1

   Load the model

   The site downloads a self-hosted face-landmark model.

2. 2

   Read the image locally

   The browser decodes the selected file into memory.

3. 3

   Calculate locally

   Landmarks, measurements and score remain on the device.

4. 4

   Discard with the tab

   The site has no image database or server upload path.

Similar spellings lead people to different products. Omoggle is the live stranger-webcam competition discussed in our safety guide. Omogle here is a private, single-photo analysis with no stranger matching, leaderboard or public result.

After a user selects a file, the browser decodes it and runs a self-hosted face-landmark model. Geometry functions use those landmarks to calculate symmetry, eye-corner angle, facial width-to-height ratio, visible jaw-angle approximation, inner-eye spacing and middle-to-lower facial balance. A fixed scoring function then combines the measurements into the PSL-style summary.

The detector estimates points, the geometry code turns them into measurements, and a fixed scoring function applies the product’s targets. The interface then shows the result. Because these parts are separate, they can be tested without asking a language model to invent a rating for each upload.

• Detector: locates visible landmarks.
• Geometry: calculates defined ratios and angles.
• Scoring: applies a fixed heuristic.
• Interface: exposes free values and explanations.
• Optional report: adds depth without rerunning the face.

The application has no upload endpoint for the selected file or facial landmarks. Network requests fetch site assets, the face model, analytics events and optional checkout services; they do not carry image bytes. Automated browser tests also fail if a scan request contains the test image data.

Cookieless analytics records product events such as route and scan completion, without adding face-derived measurements or identity. The paid report is restored through an entitlement token rather than a profile containing a photograph. Closing the page releases the in-memory image from the application flow.

• No account is required for a scan.
• No gallery of previous faces is created.
• No landmark coordinates are sent to analytics.
• No score or metric becomes an analytics property.
• No server-side face inference is performed.

Omogle does not claim a validated population percentile, medical diagnosis or universal attractiveness measurement. It does not infer race, health, personality, sexuality or social status. It does not recommend surgery from a score. Those exclusions are both ethical boundaries and technical facts: the selected photograph and six geometric inputs cannot support such conclusions.

The design may be playful, but the limits are part of the result. Reference targets are heuristic, the jaw measurement is not radiographic anatomy, and photo conditions can change the numbers.

Some AI rating products send an image to a general model and ask for an opinion, so phrasing, model updates or sampling can change the answer. Omogle uses a fixed detector and explicit geometry. There is no temperature setting, hidden panel persona or prompt asking the model to sound certain.

A fixed calculation is easier to test. Unit tests cover the geometry, content checks reject unsupported ranking claims, and browser tests exercise the scan. This does not make attractiveness objective; it simply means the same calculation is applied every time.

A future scoring or detector revision could still change results, which is why versions should be deployed deliberately and dates on explanatory pages matter. Within one deployed version, identical input follows the same tested calculation path from landmark detection through the displayed score.

Repeatability also depends on the input. The same face photographed from a different distance, under different light or with a different expression is a different image and may produce different geometry. For a useful comparison, keep the camera, distance, light and expression as similar as possible. The tool rejects obvious pose problems, but it cannot remove every camera effect.