Google Home Familiar Faces Update Uses Clothing and Body Size to ID People

Google is updating its Familiar Faces feature for Google Home cameras on June 23rd. The change lets the system identify people you have already tagged even when their face is not clearly visible, by reading additional signals such as body size and clothing color. The company is also adding automatic library updates so that saved images of household members stay current, which should cut down on false alerts triggered by old reference photos.

What happened

On June 23rd, Google is rolling out an expanded version of Familiar Faces, the facial recognition feature built into the Google Home camera platform. The core change: tagged people can now be identified even when the camera cannot get a clear view of their face.

According to Google, the system will draw on what it calls “additional non-biometric signals,” specifically body size and clothing color. So if a tagged household member walks past a camera facing away, the system has more to work with than just a clear facial image.

The second piece of the update addresses a separate problem. Familiar Faces libraries will now update automatically with the most recent images of each recognized person. Previously, if your reference photos were old, you could receive inaccurate notifications. The automatic refresh is meant to fix that.

Why it matters

Smart home cameras are only as useful as their accuracy. A system that constantly sends alerts because it cannot tell a family member from a stranger wastes your attention and trains you to ignore notifications. Both parts of this update target that exact failure mode.

The non-biometric framing is worth noting. Google is drawing a deliberate line between facial geometry data and secondary visual cues like shirt color or build. That distinction matters for privacy regulations in several jurisdictions, and it signals that Google is trying to improve accuracy without expanding the legal surface area of biometric data collection.

For households where cameras cover entryways, back doors, or stairwells where people rarely face the lens directly, this update could meaningfully reduce noise in your notification feed.

Our take

This is a practical, incremental improvement. Most people with Google Home cameras have experienced the frustration of a system that fails the moment someone carries a bag in front of their face or walks away from the lens. Using clothing color and body size as supporting signals is a sensible fallback, not a replacement for facial recognition.

That said, clothing color is inherently unreliable across a day. Someone who leaves the house in a blue jacket and returns in a white T-shirt may still confuse the system. Google has not published accuracy numbers for this combined approach, so it is worth tempering expectations until real-world results come in.

The automatic library refresh is arguably the more quietly useful change. Stale training data is a common, overlooked source of smart home camera errors, and fixing that in the background without requiring user action is exactly the kind of maintenance feature that actually gets used.

If you manage a Google Home setup for a client or for your own business premises, June 23rd is a good date to check whether Familiar Faces is enabled and whether all relevant people are properly tagged before the update rolls out.

What to do about it

• Open the Google Home app and confirm that Familiar Faces is turned on for each camera where you want person recognition.
• Review the people already tagged in your Familiar Faces library. Remove anyone who should not be recognized, and add anyone who is missing.
• After June 23rd, monitor your notification feed for a week to see whether false alerts drop. If recognition is still off, the library may need a few days to accumulate fresh reference images.
• Keep in mind that clothing-based signals will be less reliable for people who wear very similar colors day to day or who share a similar build with another household member.

Check your Familiar Faces library before June 23rd so the automatic refresh starts with accurate, up-to-date reference images.

Source: The Verge · AI