*PSL/scale
PSL ratingPricingLearnScale
Account
Home/Scoring method

Method and evidence

How PSL Scale produces a face rating.

This page describes the current service implementation. It is an operator disclosure checked against the upload and analysis code on November 16, 2026, not an independent validation of attractiveness or scoring accuracy.

From image to private report

  1. Upload and checks. A signed-in user selects a JPG, PNG, or WebP file they have permission to use. The upload interface shows the applicable file limit; the service checks file type and size before analysis. The photo guide explains usable framing and upload errors.
  2. Temporary image preparation. In the current analysis path, Cloudflare Images scales the image down to fit within 1024 × 1024 pixels and prepares a temporary JPEG in request memory for Cloudflare Workers AI. The original image is not stored by this path. The Privacy Policy explains provider processing and result retention.
  3. Model response and checks. The service asks the vision model to describe only visible presentation, not identity, health, personality, ethnicity, or other protected traits. It requires a 0–8 overall score, at least one named 0–8 dimension with an explanation, and a written summary. It checks the response structure before saving a result to private history.

What is evidence for these statements?

The file types and paid-plan limits come from the service upload validator; the resize and model request come from its current Workers AI adapter; and the score range and required explanation fields come from the result validator. The saved report displays the model identifier used for that result. These are implementation facts, not proof that the rating measures a stable property of a person. To interpret a completed report, read the rating scale guide.

What this method cannot establish

This site does not publish an independent calibration or repeatability study. It does not define population percentiles, a universal “average” score, or a fixed weighted formula for the named dimensions. The model can choose different dimension names and explanations for different images. A different photo or model response can produce a different number; see why face ratings change. Do not use this report for health, employment, dating, identity, or other decisions about a person.

Unscorable images and other failures

If the model cannot identify a clear front-facing face, it is instructed to return an Eligibility explanation with a zero input-suitability score. That is not a rating of the person. A file rejection, model error, or invalid model response is a separate failure; the page should show the error rather than present a completed score. Check your visible account history and credit state before retrying. A failed analysis does not consume a usage credit.

Corrections and updates

If a step here differs from the live service, use the content correction contact. Include this page URL, the statement to correct, and evidence or steps to reproduce it; do not email a photo, password, verification code, or card details. We will update this description when the processing path changes. Browse all face rating guides or open the tool.

*PSL/scale

A private, experimental image scoring tool for exploring facial proportions.

ExplorePSL ratingPricingAll guidesThe scalePhoto guideWhy scores varyOur method
About & policyAbout PSL ScaleContactTermsPrivacyRefunds & cancellation

(c) 2026 PSL Scale - Operated by dong, China. Address: Shandong Binzhou, China. Support: dw6179694@gmail.com.