Help & FAQ

Pick one spot and one time of day, and go back to it. The app does the rest of the work, but it cannot invent light you do not have.

• Same place, same hour. A window at 9 a.m. is a different lamp at 9 p.m. Overhead ceiling light and a window at the side produce very different shadows on the same skin.
• One light source, from the front. Face the window or the lamp rather than standing with it behind you.
• Bare face. No make-up, no just-washed shine, no just-applied cream. Give the same amount of time after washing each day.
• Follow the three gauges, not your judgement. When all four readings are in range the shutter unlocks by itself.

The flash is switched off on purpose and cannot be turned on: a flash overrides the ambient light the whole method depends on.

SameLight locks the camera’s ISO, exposure and white balance, then measures four things. All four have to be in range before the shutter arms:

• Light — ambient brightness between 250 and 800 lux.
• Colour of the light — within ±500 K of the white balance the camera locked for this session.
• Distance — your face 25 to 40 cm from the lens.
• Head angle — yaw, pitch and roll all under 6°.

Whichever axis is furthest out gets you a number, not a mood: “Move 4 cm closer to the camera.”, “Brighter by 15 percent — face a window or lamp.”, “Straighten your head by 3 degrees.”, “Light looks too warm by 850 K.” The advice is always the smallest change that would clear the gate, not a move to the middle of the range.

If a reading cannot be taken at all — no face in frame, for instance — that axis shows “Can’t be read” and the shutter stays locked. An unreadable axis is never counted as a pass.

Three gauges, four axes: light and colour temperature share the LIGHTING gauge, which always shows whichever of the two is worse.

When the gate will not arm and you want the photo anyway, there is a second button: Save anyway as baseline (not standardized). It exists because your very first photo is often taken before you have worked out where the good light in your home is, and losing that day entirely is worse than marking it honestly.

A baseline is stored like any other photo, but:

• it carries an amber dot and a BASELINE tag everywhere it appears;
• it is excluded from the trend chart — a trend needs two photos that both cleared the gate;
• when the app picks a comparison partner for you, it skips baselines.

You can still select a baseline by hand in the timeline and compare it; the comparability score then decides, exactly as it would for any other pair, and the amber tag stays visible. In the export CSV the column is_baseline marks it.

Because the two photos scored below 70 out of 100 for comparability, and drawing that comparison would have shown you a change in your lighting or your position and let you read it as a change in your skin.

You are not left with nothing. The blocked screen names the axis that was furthest apart, tells you by how much — “differs by 22 percent”, “differs by 9 degrees”, “differs by 1200 K” — and offers a retake matched to that axis, or a different photo to compare against.

This is a refusal, not a hidden result. Below the threshold the aligned crops are never computed in the first place; there is no faded version of the comparison sitting behind a paywall or a warning banner.

There is also a different failure that reads similarly: if a face cannot be located precisely enough in one of the two photos, no comparison is possible at all. The app says so in those words, because it is not the same thing as the photos being too far apart.

Four differences between the two photos, each measured, scaled and subtracted from 100:

The distance and colour weights are not tuned by feel — they are derived from the capture gate. Two photos that both cleared the gate can differ by at most 1000 K of colour and by a known maximum in face size; at exactly those worst legal points, each axis on its own uses up exactly the 30-point budget between 100 and the threshold. Passing the gate twice therefore does not guarantee a comparison: it guarantees the arithmetic is bounded.

If head pose cannot be read in either photo, the angle axis takes the full penalty rather than being treated as zero.

On screen the four axes are shown as three rows — colour temperature is folded into LIGHTING — so each visible row carries a budget of 10 points.

It is a two-minute test you run on yourself, in Settings › Tools › Stability check: five photos in a row without moving, each counted independently, then the coefficient of variation across the five counts.

Under 10 percent it passes, and the app tells you what that buys you: differences bigger than 10 percent in later photos are real change rather than noise. Over 10 percent it says the opposite — treat small differences as noise, and retake in steadier light.

What it measures is your conditions with your face and your lighting, not a number the developer measured once and printed on a box. The first run is free forever; repeat runs are part of Pro.

Photos are written to SameLight’s own storage inside the app container on your iPhone — not the system photo library, not a shared folder. Both photo directories are flagged to be excluded from iCloud and iTunes backups, so your face is not copied into a backup you forgot you had. There is no account, no server-side copy and no sync.

All analysis — alignment, spot counting, comparability scoring — runs on the device.

The one thing the app sends is an anonymous usage count: an app id that is the same for everyone, an event name from a fixed list, a random 12-character session id that is thrown away after 30 minutes of inactivity, and occasionally one short label from a fixed list. No photo, no crop, no thumbnail, no facial landmark, no spot count, no capture reading, no advertising identifier, no location. You can switch it off entirely in Settings › Privacy › Analytics. The full list is on the privacy page.

Settings › Export ZIP. The app writes a single ZIP and hands it to the system share sheet, so you choose where it goes — Files, AirDrop, a mail attachment, a cloud drive of your own choosing. Nothing is uploaded on your behalf.

Inside:

• photos/ — every original JPEG, stored uncompressed inside the archive.
• capture-params.csv — one row per photo: id, date, iso, shutter_ms, wb_kelvin, ambient_color_temp_k, wb_source, distance_cm, distance_source, yaw, pitch, roll, angle_source, ambient_lux, face_box_ratio, is_baseline, spot_count. Readings that could not be taken are left empty, never written as zero.
• comparisons.pdf — one page for each comparison you actually saw in the app.
• README.txt — what each column means and the scoring formula with its constants, so a third party can recompute the numbers.

One switch, Include capture metadata, controls both the camera columns in the CSV and whether EXIF stays in the exported JPEGs. Turning it off gives you a plainer, less identifying archive — and, as the app says on that switch, an export that can no longer be recomputed. The pixels are copied bit-for-bit either way; nothing is re-encoded.

Export is free forever and always covers your whole history, including the photos Pro would put back on the timeline.

A single photo can be deleted from its own detail screen. Everything at once is Settings › Your data › Delete all data, which removes every photo, capture parameter, comparison and stability check from the device and tells you how many records and files went.

It cannot be undone and we cannot help you recover anything, because there is no copy anywhere else — that is the same property that keeps your photos private. Export first if there is any chance you will want them.

Deleting the app removes the same data.

Free forever: standardized capture and the gate, on-device spot counting, comparison with the comparability score, the complete export, and deleting your data. Also your first stability check.

The first 14 days: everything is unlocked, including the Pro parts. Inside that window the app is structurally unable to present a full-screen upgrade prompt, and Settings shows how many free days are left with no upgrade button anywhere on the page.

Pro, after day 14: the timeline beyond the rolling 14-day window, the trend chart, and re-running the stability check. $4.99 per month or $29.99 per year, with a 3-day free trial on the yearly plan. Nothing is deleted when the free window ends — your older photos stay on the device and stay in every export; Pro puts them back on the timeline.

Cancel in your Apple Account settings, at least 24 hours before the period ends. The app opens Apple’s own subscription sheet for this; it does not send you to a web page. Restore is in Settings › Subscription › Restore purchase and also on the paywall itself, visible whether or not you are subscribed. Refunds are handled by Apple, not by us.

iPhone running iOS 17 or later, portrait orientation. There is no iPad layout.

On iPhones with a TrueDepth front camera the distance reading comes from the depth sensor. On iPhones without one, the distance is estimated from the width of your face in the frame; the app labels those readings estimated on screen and records the difference in the CSV column distance_source, so you always know which kind of measurement you are looking at.

The interface, the retake advice and the bundled privacy, terms and method documents are all available in English, Simplified Chinese, Spanish, French, Brazilian Portuguese and Japanese.

Two separate questions, and only one of them has an answer today.

Stability — does the same face under slightly different conditions give the same number? That is what the counter is built and tested for: against controlled brightness, white-balance, rotation, translation and compression changes of the same photo, and against your own five-shot stability check on your own face.

Absolute accuracy — would a person counting by hand agree with the number? That is still being calibrated. On the current test images the counter tends to over-count faint redness, and no agreement figure is published because none has been earned. This is why the app never grades, never scores your skin and never puts a single number in front of you as a verdict.

The practical advice: read the direction of the trend across many standardized photos. Do not read one number, and do not read two photos that the app told you were not comparable.