How to choose the best photo for face search
A better query photo cannot guarantee a match, but one clear adult face with useful lighting and limited obstruction gives the detector and search provider cleaner visual information.
Face search begins with the visual information available in your query. A strong photo helps the local detector find the face and gives the provider a cleaner comparison. It does not create coverage where no indexed source exists, and it does not remove the need to verify results.
Use exactly one adult face
Group photos create ambiguity about which person should be searched. FindFace rejects images with multiple detected faces rather than guessing. Choose a portrait or crop where one adult face is clearly dominant.
Make the face large enough
A tiny face contains fewer usable pixels and more compression artifacts. Avoid screenshots where the portrait occupies only a small profile badge. When possible, use the original image or a higher-resolution copy rather than enlarging a tiny thumbnail.
Prefer natural, even lighting
Strong backlighting, crushed shadows, colored nightclub lighting, and blown highlights can hide useful facial structure. Soft daylight or balanced indoor light is usually easier to analyze. Normal shadows are fine; the goal is visible detail, not a studio portrait.
Choose a realistic angle
A front-facing or slight-angle photo is a reliable starting point. Extreme side profiles, tilted faces, very wide-angle distortion, or expressions that hide much of the face can make detection and comparison harder. If you have authorization to use more than one photo, try a second natural angle rather than repeatedly editing the same weak image.
Limit obstruction and filters
- Avoid sunglasses that hide both eyes.
- Avoid masks, hands, hair, or objects covering a large part of the face.
- Avoid beauty filters that substantially reshape features or skin texture.
- Avoid stickers, text, watermarks, and heavy sharpening across the face.
- Avoid motion blur and aggressive screenshot compression.
Do not upload more context than the search needs
FindFace analyzes the original file locally, detects one face, generates a standardized crop, and re-encodes that crop before a confirmed upload. Re-encoding removes metadata from the original file. The original image is not needed as a server-side search record.
You should still avoid images containing documents, children, private interiors, addresses, computer screens, or unrelated people. Use only a photo you are permitted to search under the product’s responsible-use rules.
If the first search is weak
- Check whether the face was small, blurred, obstructed, or strongly filtered.
- Try a legitimate second photo with a more natural angle or clearer lighting.
- Use exact-image search if your real goal is finding copies of that specific image.
- Accept that the relevant public page may not exist in the provider’s current index.
Better input can improve the query, but it cannot make a visual score an identity verdict. Review the complete reverse face search workflow before interpreting a candidate.