The best photo for a face search

What makes a photo match well

• One face. The engine works on a single face at a time. Crop group shots down to the person you are searching for.
• Front-facing or three-quarter view. The description it builds relies on seeing both eyes, the nose and the mouth.
• Even light. Daylight or soft indoor light. Harsh side light, backlight and heavy shadow hide features.
• Enough pixels. A face a few hundred pixels across is comfortable. Below roughly 60 pixels across, matches become unlikely.
• Natural appearance. No beauty filters, face-altering effects or heavy makeup changes. Glasses are fine; sunglasses are not.
• Sharp focus. Motion blur and soft focus remove the detail the comparison depends on.

What makes a photo match badly

• Full side profiles. Half the face is missing from the measurement.
• Sunglasses, masks, hands or hair across the face. The eyes and mouth carry a lot of the signal.
• Very small faces, such as a person in the background of a wide shot or a tiny avatar.
• Heavy filters and AI retouching, which change the geometry the engine measures.
• Low-resolution screenshots of screenshots, where compression has smeared the detail.
• Very old or very young photos when you are looking for recent pages. Faces change; a current photo finds current pages better.

Formats and size

FaceSearch accepts JPEG, PNG and WEBP. Large photos are downscaled in your browser before upload, so a full-size phone photo is fine as it is; you do not need to shrink it first. Screenshots are usually PNG and work well once cropped. If an image is refused, re-save it in one of the three formats and try again. HEIC photos from iPhones are converted when you share them to the browser on most devices; if not, export as JPEG first.

How to prepare a screenshot

1. 1

   Take the screenshot at full size

   Open the photo as large as the app allows before you capture it. A screenshot of a thumbnail has too few pixels on the face.

2. 2

   Crop to the face

   Use your phone's built-in editor or any image tool. Cut away the app interface, names, captions and other people. The face should fill most of the frame, with a little space around it.

3. 3

   Do not zoom past the pixels

   Cropping removes surroundings; it does not add detail. If the face is tiny in the original, look for a bigger photo rather than enlarging a small one.

4. 4

   Upload the crop

   Drop it on the search page. If the result is thin, try another photo of the same person rather than the same crop again.

Searching with several photos

FaceSearch searches one photo at a time, and that is a feature when you use it well. Different photos of the same person emphasise different features, so a front-facing shot and a three-quarter shot can surface different pages. Run them one after another and compare the lists. A page that appears for both photos is a stronger lead than one that appears for only one; a look-alike usually matches one photo and vanishes on the next. Each search uses a credit, so pick your best two or three rather than everything you have.

Why a good photo can still return nothing

• The face is not on any public page. Most personal photos live inside private accounts, which no face search engine can see.
• The pages exist but have not been indexed yet. The index follows the web with a delay. A later search can find what today's did not.
• The public photos look very different. A decade-old passport photo and today's pages may measure far apart. Try a photo from the period you are interested in.
• The face is AI-generated. It has no original to find.
• A wider search would help. A deep search covers more public sources than a normal one.