Face Search

Face search helps you find photos of the same person or visually similar faces in an indexed library.

Steps

1. Click Face in the top-right workspace toolbar.
2. Upload, drag, or paste a reference image that contains a face.
3. Wait for face detection.
4. Select the target face if multiple faces are detected.
5. Set face similarity, the maximum result count and scope: the whole database, selected folders or selected collections. Then start searching.
6. Review the returned images.

The results open in the Face Search Results drawer and remain in the Results sidebar history. A single detected face is selected automatically; choose the intended face manually in a group photo.

Face Search or Face Filters?

Face Search uses a reference face. Face Filters needs no reference image and finds photos by face count, age, gender, smile, glasses, beard or face area. Both require face data; an older library missing attributes may need Face Database repair before filtering.

Reference Image Suggestions

Use a clear face whenever possible:

• front-facing or near-front-facing;
• not too small in the image;
• not heavily blocked by sunglasses, masks, or motion blur;
• enough lighting and resolution.

Common Cases

• If no face is detected, try a clearer image or crop the face area.
• If results contain similar-looking people, narrow the scope or use a better reference face.
• If face search is unavailable, generate or rebuild face data first.
• If too few results return, lower face similarity or expand the scope. Review the matches manually.

RAW search inputs

From 3.13.27, image, reference-face and pose search accept CR2, CR3, NEF, ARW, DNG, RAF, ORF, RW2 and PEF through file selection, drag-and-drop and file paste. Reading a reference file does not import your library: prepare its matching search data first. Copying a path as text is not the same as copying the image file.

Screenshot Example

The public sample portrait contains one detected face, selected automatically. Searching the prepared library returns the portrait and its identical backup at 100%. Before a group is named, the result can show photos while the People count remains zero. This demonstrates detection and retrieval of identical images; it is not an accuracy claim for different photos or an identity verification.

Sample image: NASA public-domain astronaut image distributed in the scikit-image sample dataset.