What a Face Report Actually Gives You (And What It Doesn't)

aiportrait — Wed, 07 Oct 2026 06:30:43 +0000

A face report is not a verdict on your face. It is a structured read on your facial proportions — one harmony score, four measurements behind it, a face shape, and a visible symmetry overlay. Here is what each of those is genuinely useful for, and where the tool stops being relevant.
Most face-scoring tools have a problem: they give you a number and stop there. You get a 7, or an 82, or a tier label, and then you stare at it with no idea what to do next. The number is the easy part. The useful part is knowing which measurement produced it — because that is the part you can actually act on.
That is the gap a face report is built to close. It takes one photo and returns four separate measurements, a face shape with percentage matches, and an overlay that makes your left-right difference visible instead of leaving it as an abstraction.
Below is what each of those outputs is actually good for — including the cases where you should ignore it entirely.
Illustration of a face split by a vertical mirror axis, with four measurement dials around it

The Four Measurements, and What Each One Is For

A single 0-100 harmony score is the least informative output on its own. It compresses four genuinely different things into one number, and most of the value sits in the breakdown.
The full breakdown is on the face analysis page, which explains how each dimension is scored against a reference range. In short, the four are:
Symmetry — how closely the left and right sides of your face match. Read this one when you want to see the direction and size of any left-right difference. The face report does not just score it, it draws it: your detected facial contour is overlaid with its mirrored counterpart directly on your photo, so you see the asymmetry rather than infer it from a number.
Facial thirds — how the face divides vertically into forehead, midface, and lower face. Read this one when you are thinking about proportion balance, or about hair and glasses that change how your face reads vertically.
Facial fifths — how the face divides horizontally across its width. This is the one stylists reference most often when reasoning about eyewear width against face width.
Golden ratio alignment — how closely your proportions track classical ratio conventions. Read this as a curiosity rather than a target. It is the most culturally loaded of the four, and a low score here is not a defect.
Tip: Read the four scores, not the total. A high overall number with one weak dimension tells you something useful; a mediocre overall number with four balanced dimensions tells you something different and equally useful.
If you want only the single number and none of the breakdown, the face rater reports the same 0-100 harmony score on its own. If you already know your score and want to understand why it landed there, start with the full report instead.

Use Case 1: Settling an Unexplained Photo Bias

The problem. You have a photo where you look worse than you expected, and you cannot say why. Side profile, camera angle, lighting — something shifted, and you have no way to isolate it.
How the report helps. Run it and read the symmetry overlay. Most of the time the visual problem is a left-right difference that only becomes visible once the mirror contour is drawn over your photo. Seeing that a specific region pulls left is more actionable than being told the overall number is lower than you expected.
Steps.

Upload the photo in question.
Open the symmetry overlay and look at where the two contours diverge.
Compare against a second, more neutral photo of the same day.
Note whether the difference follows the photo or stays constant. If you have never done this before, the fastest way to get comfortable with it is to see the overlay on a photo you already know well. Try the face report free on one photo, then compare it against a second one before drawing any conclusions. The value. Step 4 is the actual payoff. A difference that shows up in one photo and not the other is a photo artifact. A difference that shows up in both is geometry. Most people cannot tell which they are looking at, and this separates them in one glance. Diagram showing a face outline overlaid with its mirrored dashed counterpart, with arrows marking the diverging areas

Use Case 2: Choosing Eyewear That Actually Fits

The problem. Glasses frames are sold by a number on a card, and the number means nothing to you. You try frames in a shop, the shop is not lit properly, and you end up buying something that looks wrong in a photo two days later.
How it helps. Read your facial fifths score and your face shape first. Facial fifths tells you how your face divides across its width — the dimension eyewear width is actually chosen against. Face shape tells you what you are working with to begin with.
Steps.

Get your face shape from the face shape detector, including the three closest matches with percentages.
Read your facial fifths score in the face report.
Use those two together as a starting frame recommendation — narrow frames for faces that read wider than they are long, the inverse for longer faces.
Confirm in person, under lighting close to how you will actually be photographed. The value. This turns frame selection from guesswork into a two-input decision. And the percentages matter more here than the top label. When Oval and Round both sit near the top, that tells you the frame direction is genuinely ambiguous — which is worth knowing before you spend money, not after.

Use Case 3: Understanding Why Your Score Moved

The problem. You ran a face rating before and got a different number, and now you do not trust the tool. This is the most common reason people abandon face scoring tools entirely.
How it helps. Two things. First, every result comes from detected facial landmarks, so the same photo always returns the same score — the process is deterministic, not a roll of the dice. Second, the four dimension scores tell you what changed rather than just that something did.

Steps.

Run the rating on the original photo and note the four dimension scores.
Run it on the second photo.
Diff the dimensions, not the headline number.
Check the input conditions before you blame the tool. The value. A point or two of overall movement almost always resolves into one dimension moving, usually symmetry, usually from head angle in the photo. That is a finding about your photo, not about your face — and it is the difference between a tool you trust with a varying input and a tool you write off entirely. Filters, hard shadows, and tilted angles all shift detected landmarks. That is why the face rater treats a stable, front-facing, evenly lit input as the baseline for a repeatable number, rather than presenting a score as a fixed property of your face.

Use Case 4: Checking Whether a Photo Is Even Usable

The problem. You want a reliable read but you have several photos and no idea which one to use. A bad input gives a bad measurement, and you would rather not chase a number you cannot trust.
How it helps. Run the tool on two or three candidates and compare the dimension scores. A photo that produces a wildly different profile from the others is the bad input, and it will usually be obvious in the symmetry score and the overlay before you look at anything else.
Steps.

Pick three candidates: straight-on, evenly lit, whole face visible, hair clear of the forehead and jaw.
Run each one.
Discard the outlier.
Keep the two that agree with each other. The value. Agreement between two photos is your confidence signal, and confirming it costs you nothing. This is also the fastest route to a stable reading, because it removes the single largest source of variance before you start interpreting anything.

Use Case 5: Getting a Shape Label Without the Full Report

The problem. You do not need a score at all. You need to know your face shape for a haircut decision, a makeup placement question, or a pair of glasses — and you do not want to read a four-part breakdown to get it.
How it helps. The face shape detector returns the answer on its own: one closest shape, plus the next two closest with a percentage each, across six shapes — Oval, Round, Square, Heart, Diamond, and Long.
The six detected face shapes: oval, square, round, diamond, heart, and oblong

Steps.

Upload a clear, front-facing photo.
Read the top shape and its percentage.
Read the two runners-up — the gap between first and second is the actual information.
Treat the shape as one input among several for your styling decision. The value. It is the one output that translates directly into a real-world choice, which is why it is worth separating from the score. Face shape is a practical reference for glasses, hair, and grooming, and unlike the harmony score it does not need interpreting before use. It is also worth knowing that two people can look quite different and still return the same shape, because the label comes from measured proportions rather than visual impression. That is precisely what makes it a reliable input rather than a judgment — and it is why the percentage gaps tell you more than the winning label does.

Where This Stops Being Useful

Being straight about the limits is more useful than a longer feature list.
It is not a beauty score. The harmony score is a geometric measurement, not a validated attractiveness metric and not clinical anthropometry. Read it as a description of proportion balance and nothing more. If a score of 74 makes you feel bad, the tool is not working as intended — the number was never measuring that.
It is not medical. No diagnosis, and no dermatological, orthodontic, or cosmetic advice. If you are asking whether something is medically wrong with your face, this is the wrong tool.
It is not personality. No trait, mood, or intelligence is read from the geometry of a face. Any tool claiming to extract those from a photo is doing something else entirely.
Percentiles are not other people. Where a reference percentile appears, it comes from a reference distribution, not from a comparison against real users. It is a positioning aid, not a ranking.
One photo is a snapshot. It reflects lighting, angle, and expression as much as it reflects geometry. This is exactly why Use Case 3 exists.

And Why It Runs in Your Browser

Worth calling out separately, because it is a real constraint rather than a feature bullet: the analysis runs locally. Your photo is never uploaded to a server, and no copy is created. When you generate a share card, the card carries your score and tier — not the original photo file, and no original location data.
That matters most in the exact cases above. Use Case 4 means uploading several photos of yourself; Use Case 2 and Use Case 3 mean you may be using photos that were never publicly posted. Being able to run those entirely on-device removes the reason many people hesitate to try a face tool at all.

Getting the Most From One Photo

Input quality dominates everything. Straight-on, evenly lit, whole face visible, hair clear of the forehead and jaw. This single factor affects the result more than any other.
Run two photos, not one. Agreement between two inputs is your confidence signal, and it is free.
Read dimensions before the total. The breakdown is the output; the total is only a summary of it.
Diff dimensions, not headline numbers. When a score changes between photos, the dimension that moved is the actual finding.
Use the percentage gaps. Whether for face shape matches or reference positioning, the gap between values tells you more than either value alone.
Re-run after a real change. The measurement is a snapshot, so a new haircut or a new photo setup is worth a re-check.

FAQ

What does a face report actually measure?

It measures facial proportions from one photo: bilateral symmetry, facial thirds, facial fifths, and golden ratio alignment. Those four are scored separately and then summarized into a single 0-100 harmony score with a tier label. Nothing outside the geometry of the photo is measured.

Is a face report a beauty score?

No. It is a geometric measurement reference, not a validated attractiveness metric and not clinical anthropometry. It scores how your proportions balance, and it does not judge your appearance.

Why does my score differ between photos?

Most often one dimension moved, usually symmetry, usually because of head angle, lighting, or expression. Because every measurement comes from detected landmarks, the same photo always returns the same score — so a changed result points at the input, not at an unreliable tool.

Does the face report upload or store my photo?

No. Processing happens entirely in your browser, so the photo is never uploaded to a server and no copy is ever created. A shared report card carries your score and tier rather than the original photo file.

Do I need an account to use it?

No. There is no sign-up, no email step, and no payment. Open the page, add one photo, and read the result.

What face shapes can it detect?

Six: Oval, Round, Square, Heart, Diamond, and Long (also called oblong). You get the closest match plus the next two closest, each with a percentage showing how near the match was.

What should I do if my score is lower than I expected?

Start with the four dimension scores rather than the headline number. They show which specific measurement pulled the total down, and photo conditions such as light, angle, and expression often explain part of the gap.

The Whole Picture, In One Place

The full report — harmony score, all four dimensions, face shape with matches, and the symmetry overlay — is on the FaceReport home page. Each measurement also has its own page if you only need one: face analysis for the four dimensions, face rating for the single score, and the face shape detector for the shape label alone.
Free, no account, and your photo never leaves your device. Try the full report free on one photo, then read the four dimensions instead of the headline number. If you would rather start with a single number, get a free face rating first and come back to the breakdown once you know the number you are trying to explain.
The most useful thing you can do with a face report is not read the score. It is find out which measurement moved it — and then go look at your own photo again with that in mind.
Disclosure: This article is written by the FaceReport team about our own product. Every capability described here — the four dimension scores, the symmetry overlay, the six-shape classification with percentage matches, and local browser-based processing — reflects features that run today on the linked pages, Tested against the live product and its own published output, and Editorially assessed against the site's own published documentation rather than against third-party benchmarks.
This was a hands-on feature test of our own tooling, not a statistical accuracy benchmark: no claim is made here about clinical accuracy, and no comparison against other products was performed. No third-party sites are linked or recommended in this article, and this is not a comparative review. Screenshots and numbers shown are geometric measurements of the kind of output the tool produces, and are reference points rather than clinical measurements.
This article describes geometric measurement features of FaceReport for informational purposes. Results are reference points, not clinical measurements, diagnoses, or judgments of appearance. Intended for users aged 16 and over.

Show HN: AI Upscaler – Free, no-signup image enhancer up to 4x with strict privacy

aiportrait — Sat, 30 May 2026 15:30:51 +0000

Hi HN,

I built AI Upscaler , a free online tool to enhance and upscale images up to 4x without losing quality.

There are plenty of upscalers out there, but many are locked behind heavy paywalls, require annoying sign-ups, or have questionable privacy policies. I wanted to build something fast, accessible, and respectful of user data.

Key Features:

No Sign-up / Free: You can use it instantly without creating an account.

Privacy-First: All uploaded images are processed via HTTPS and automatically deleted from our servers after 60 minutes. We have a strict zero-training policy (your images will never be used to train models).

Multiple AI Models: Optimized for different use cases including Balanced, Pro, Graphics, and text enhancement.

Face & Old Photo Restoration: Built-in face enhancement to fix blurry portraits or vintage photos.

It supports PNG, JPG, and WebP up to 10MB.

I would love to get your feedback on the upscaling quality, processing speed, and UI/UX. What other specific models or features (e.g., batch processing) would you like to see next?

Thanks!

DEV Community: aiportrait

What a Face Report Actually Gives You (And What It Doesn't)

The Four Measurements, and What Each One Is For

Use Case 1: Settling an Unexplained Photo Bias

Use Case 2: Choosing Eyewear That Actually Fits

Use Case 3: Understanding Why Your Score Moved

Use Case 4: Checking Whether a Photo Is Even Usable

Use Case 5: Getting a Shape Label Without the Full Report

Steps.

Where This Stops Being Useful

And Why It Runs in Your Browser

Getting the Most From One Photo

FAQ

What does a face report actually measure?

Is a face report a beauty score?

Why does my score differ between photos?

Does the face report upload or store my photo?

Do I need an account to use it?

What face shapes can it detect?

What should I do if my score is lower than I expected?

The Whole Picture, In One Place

Show HN: AI Upscaler – Free, no-signup image enhancer up to 4x with strict privacy