This video introduces "Google AI Edge Gallery," a new, experimental, open-source application that allows users to run artificial intelligence models locally on mobile devices. Currently available for Android via an APK on GitHub (with an iOS version planned), the app is not yet on the Google Play Store. It serves as a platform for on-device machine learning, enabling users to download and interact with various AI models, such as Google's Gemma 3, which are hosted on Hugging Face. A key feature highlighted is the ability to run these models using either the device's CPU or GPU, with a setting to switch between them for performance tuning. The application also supports the import of custom models, making it a useful tool for developers and AI enthusiasts wanting to test models without cloud dependency.
Google's New App for On-Device AI: An Overview of AI Edge Gallery
00:00:02 Introducing AI Edge Gallery: Google has released a new application that allows AI models to run locally on mobile devices, utilizing the phone's own hardware instead of the cloud.
00:00:15 Experimental and Open-Source: The app, named "Google AI Edge Gallery," is free, open-source, and currently available on GitHub. It is not yet listed on the Google Play Store, suggesting it's in an early, experimental phase.
00:00:57 Android First, iOS Coming Soon: The application is currently available for Android users, who can download the APK file directly from the "releases" section of the GitHub page. An iOS version is planned for the future.
00:01:25 Downloading and Managing Models: Users can download various models, such as Gemma 3 (approx. 550 MB), from within the app. These models are hosted on Hugging Face, which requires users to sign in and agree to terms. The app also allows for the import of custom models.
00:02:55 On-Device Processing with CPU or GPU: A key feature is that models run directly on the device. The app includes a setting that allows the user to choose whether to run the models on the phone's CPU or GPU, with the GPU likely offering faster performance.
00:03:33 A Tool for AI Enthusiasts: This application is positioned as a valuable tool for anyone interested in AI, providing a straightforward way to test and experiment with different models locally for various purposes.