Skip to main content
The People feature analyzes the photos and videos in your catalog and identifies every face it finds. It then groups those faces by visual similarity into clusters: collections of detections that likely belong to the same individual. You can review each cluster and give it a name, turning it into a named Person that you can browse and search across your entire media library.
People is currently in Beta. The feature is functional and available to all users, but the clustering quality and interface may change as we continue to improve it. We welcome feedback from teams actively using it.

​
How Facial Detection Works

Before faces can be clustered, 1Archive must first scan a source for faces. You trigger this by selecting Detect People on a specific source. 1Archive then analyzes every photo and video in that source, locates faces within each file, and records the position and a numerical representation of each detected face. For videos, face detection does not process every frame, it samples representative frames to keep the process efficient while still capturing faces that appear throughout the clip.
Detect People runs as a background process. You can continue working in 1Archive while detection is in progress. The more media a source contains, the longer detection will take.

​
Face Clusters

After detection completes, the individual face detections need to be grouped. Running Cluster Faces tells 1Archive to examine all detected faces that haven’t been assigned to a cluster yet and group the ones that look alike. Clustering uses the numerical face representations created during detection to measure similarity between faces. Faces that are sufficiently similar to each other, and sufficiently distinct from other groups, are placed in the same cluster. Each cluster is represented by a cover image: the highest-quality detection from that group. A few important things to understand about clusters:
  • A cluster is not the same as a Person. A cluster is an automatically generated grouping. A Person is a cluster (or a set of merged clusters) that you have named.
  • Clusters require a minimum number of detections to form. Small groups of visually similar faces may not generate a cluster if there aren’t enough confident examples.
  • New detections are incorporated incrementally. If you run Detect People on additional sources and then run Cluster Faces again, new detections are matched against existing clusters where possible before new clusters are created.

​
Named Persons

Once clusters appear, you can review them and assign names. Selecting a cluster and clicking Create Person lets you give that cluster a name, for example, Sarah or Director. From that point on, the cluster appears as a named Person in your library. If the same individual appears across multiple clusters (which can happen when images vary significantly in angle, lighting, or quality), you can merge those clusters under one Person using Add to Person.

Create Person

Assign a name to a cluster to create a new Person entry. The cluster’s cover image becomes the Person’s representative photo.

Add to Person

Merge an unnamed cluster into an existing Person. Use this when the same individual has been split across multiple clusters.
You can also Change Cover on any Person to select a different face detection as the representative image, useful if the automatically chosen cover is not the most recognizable photo of that individual.

​
Working with People Across Sources

Face detection and clustering work across all scanned sources in your organization. A Person named Sarah will surface every appearance of that individual across every drive where Detect People has been run, regardless of which source the original files live on.
Facial recognition is only available for sources that have been indexed with face detection enabled. Sources scanned without this option will not contribute faces to People. Re-running Detect People on an existing source adds detection data without requiring a full rescan.