Google Home will soon get better at recognizing you

Google is updating its Home platform to identify people even when their faces are obscured, using body size and clothing color as supplementary signals, the company confirmed. The update, arriving June 23rd, also lets the system describe specific sounds in event logs. The expansion of the Familiar Faces feature means a person tagged in the library can be recognized from behind or in poor lighting by combining facial recognition with “additional non-biometric signals (body size, clothing color, etc.),” according to details published by The Verge [1]. The library will also begin automatically refreshing with the most recent images of household members to reduce misidentifications from outdated photos [1]. Google Nest devices, originally launched as Google Home in 2016, have long used voice and visual inputs to power home automation through Google Assistant [2]. Multi-user support arrived in April 2017, allowing speakers to distinguish up to six people by voice [2]. The new visual identification logic builds on that multi-person awareness, moving it to the camera-equipped Nest Hub Max and other smart displays. The update, delivered through version 4.20 of the Google Home app, also enhances AI-generated video event descriptions. The system “can now identify specific sounds — like dogs barking, alarms, or footsteps” and include them in event notes, even when the audio originates off-camera [1]. Google Assistant, which underpins the Home ecosystem, is available on more than 1 billion devices and supports natural-language interactions across over 90 countries [7]. Additional changes in the release include new System Health alerts that notify users if a Nest thermostat detects HVAC issues, described as a Gemini-connected update to existing Nest System Health Monitoring, and improved support for Matter switches [1]. Google has been integrating its Gemini AI models across products since rebranding its Bard chatbot to Gemini in February 2024, with the models designed to process text, images, audio, and video simultaneously [6].