Google Home is getting an update that changes how its smart home AI understands what cameras see and hear. The biggest shift is in Familiar Faces, where Google is adding more context so a person can still be recognized when a camera does not get a clear look at their face.
Recognition will use more than faces
Starting June 23rd, Google is expanding its facial recognition feature for people who have already been tagged in a Familiar Faces library. The goal is to reduce cases where a smart home camera misidentifies someone simply because they are turned away from the lens.
The new system can use “additional non-biometric signals (body size, clothing color, etc.).” That means Google Home will not rely only on a clear face when trying to identify someone already known to the system.
The practical effect is straightforward: if a person in the Familiar Faces library walks through a camera view without facing the camera, Google Home may still be able to identify them. The source describes this as a way to make cameras less likely to confuse one person for another when the face is not clearly visible.
This does not describe a new group of people being identified from scratch. The update applies to people users have tagged in the Familiar Faces library, and the added signals are being used to continue recognition when the face is unclear.
Familiar Faces will refresh itself
Google is also changing how the Familiar Faces library stays current. The library will begin automatically updating with the most recent images of everyone in the house.
That matters because old examples can lead to bad notifications. If the reference images in a library no longer reflect what people currently look like, the system has less useful information to work with. By updating the library with newer images, Google says users should see fewer inaccurate notifications caused by outdated examples.
Together, the two Familiar Faces changes point in the same direction. Google Home is trying to make recognition less brittle: less dependent on a perfect face shot, and less dependent on older saved images that may no longer be the best reference.
Video descriptions will listen for specific sounds
The update is not limited to people recognition. Google also says its AI-generated video event descriptions can now identify specific sounds and include them in the notes.
The examples given are dogs barking, alarms, and footsteps. Google says these sounds can be included even when the audio comes from something off camera.
That expands what an event description can communicate. A camera clip may show one thing while the audio suggests something else nearby. If the system can label a sound that is not visible in the frame, the event note can become more useful than a description based only on what the camera sees.
The source connects these updates to issues noticed when The Verge’s Jennifer Pattison Tuohy tried Google’s updated smart home system last year. Those issues included event logs with detailed descriptions of people who were not there, and things that did not happen.
The new sound recognition and improved Familiar Faces behavior could help address some of those quirks. The source does not say they eliminate every error, but it frames the changes as steps toward better event logs and more accurate notifications.
Google Home app version 4.20 adds more smart home updates
A new update to the Google Home app in version 4.20 includes additional changes beyond camera intelligence. One is a new set of “System Health alerts” tied to Nest thermostats.
These alerts appear when a Nest thermostat detects issues with HVAC. The source describes the feature as sounding like a Gemini-connected update to the previously existing Google Nest System Health Monitoring.
The same version also brings improved support for Matter switches. The source does not provide more detail about what changes for Matter switches, so the clearest takeaway is simply that support is being improved as part of the app update.
Why this update matters for smart home AI
Smart home cameras are most useful when their alerts are specific enough to trust. A vague motion notification is one thing; a notification that names the wrong person or describes the wrong event can create confusion.
Google’s update focuses on that problem from several angles. Familiar Faces gets more current reference images. Recognition can lean on body size and clothing color when a known person’s face is not clearly visible. Video event descriptions can add sound details such as dogs barking, alarms, or footsteps.
Each change gives Google Home more context. The camera view still matters, but the system is being adjusted to interpret scenes with more than a single visual cue.
For users, the promise is fewer inaccurate notifications and richer event notes. For Google Home, the update is another step toward a smart home system that explains what happened in a home with more detail, while trying to avoid the mistaken descriptions that have made earlier AI-generated logs feel unreliable.