Score stability
Because the result describes visible relationships in one image. A different photo, a different visible state, or an updated analysis method can change the output without proving that the person changed.
Visible relationships and ratios
One-photo result
≈ range
The result describes geometry in this image, not a fixed property of a person.
Separate the causes
Before interpreting movement in a range or geometry bar, identify what changed upstream. These categories require different responses.
The person may be unchanged, but the two-dimensional projection or the visible pixels differ.
Recreate the baseline capture before comparing.
A real difference may be present at the time of capture, but one photo cannot tell you whether it is lasting.
Hold the camera conditions constant, then repeat over time.
The product or analysis approach may have been updated since the earlier result.
Create a fresh baseline after a meaningful product update.
Where variation enters
A small shift in a visible facial boundary can alter several photo proportions. The result may move even when the photograph still looks acceptable at first glance.
Standardized retest
Use this when comparing dates. The goal is not to create a flattering portrait; it is to keep the measurement conditions boring and constant.
Use the same phone or camera and the same rear/front lens. Disable ultrawide, beauty mode, filters, and portrait reshaping.
Record camera-to-face distance, lens height, phone orientation, and where you stand. Tape marks on the floor work.
Use the same broad light direction and similar exposure. Avoid mixing daylight on one date with hard overhead light on another.
Rest for a moment, relax the jaw and brow, close lips gently, open eyes naturally, and face the lens without tilt.
Keep hair, glasses, makeup, and facial hair consistent where possible. Include the same full forehead-to-chin region.
Upload the camera original, not a screenshot, messaging-app copy, or edited export. Keep orientation and crop consistent.
Take two clear captures without moving the setup. If they disagree in pattern, call the session inconclusive and inspect the setup — do not select the higher result.
Use the same tool and create a new baseline after a meaningful product update.
Do not change your appearance or repeat uploads just to optimize the output. This protocol is for comparison, not score hunting.
Interpretation
Use a chain of evidence, not one number. Stop at the first “no” and repeat under controlled conditions.
Yes · Continue.
No · Start a new baseline; the results are not directly comparable.
Yes · Continue.
No · Correct the capture mismatch and retest.
Yes · The signal is more credible, but still photo-based.
No · Treat the session as unstable or inconclusive.
Yes · Describe it cautiously as a repeated photo measurement change.
No · Do not infer that the person changed from the score alone.
Stop
Different lens, pose, light, crop, expression, or tool. The photo process changed first.
Check
Worth repeating, but one image is not enough to separate normal capture variation from a lasting change.
Stronger signal
Matched retests move in the same direction, the relevant details agree, and there is a plausible visible change.
Evidence and limits
Research supports the idea that photographs and capture conditions vary. It does not turn an informal PSL range into an objective measure of a person.
Todorov and Porter found that different photographs of the same person produced large variation in social judgments. This is perception research, not a PSL validation study.
Read sourceA 2024 study measured midface distortion at short selfie distances using both smartphones and full-frame cameras.
Read sourceICAO’s technical portrait-quality report specifies frontal pose, controlled expression, visible eyes, and nearly symmetric illumination for machine-readable portraits.
Read sourceOfficial photo guidance recommends a direct pose, neutral expression, even lighting, clear focus, and an original unedited image.
Read sourceThese references explain image variability and portrait capture. They do not validate this site’s aesthetic mapping or provide an accuracy claim.
Start with the input
Use the photo guide to create a baseline, then take the free test once and keep its limits in view.
A changed result is a question to investigate, not a verdict about you.