Almost every smartphone brand promoting AI features uses a mix of on-device and cloud-based processing, depending on feature requirements and the device’s performance. Unfortunately, not every company clearly delineates the AI tools running securely on your device from the ones that are processed off-device. Any time your personal data leaves your device and is processed on cloud servers, there is a privacy and security risk. It’s entirely reasonable to only want to use the AI features that are handled by your smartphone’s own processor as a result.
Google Pixel phones have a robust set of features powered by Google AI and Gemini, but only some of them run on-device. With Gemini Nano, your Pixel phone uses its custom Tensor system-on-a-chip to handle certain AI processing tasks. If you want to try the Pixel features that utilize the speed, privacy, and simplicity of on-device processingthese are the four most helpful ones. Aside from being more secure, they’re also available without a Wi-Fi or data connection, ensuring you have constant access to these features.
Call screen
Table of Contents
Thwart spam and scams before you answer a call
Ever wonder what it’d be like to have a personal assistant to answer phone calls on your behalf to weed out the spam and junk? That’s essentially what Call Screen does to protect you from wasting time on spam and scams, but instead of a human operator, it’s artificial intelligence sifting through your calls. Call Screen can use AI to ask who’s calling, giving users more information before deciding whether to answer the phone, block the number calling, or send the call to voicemail. With Call Screen, you never have to feel pressure to answer a call from an unknown number — even if you are expecting one.
To protect your privacy and ensure quick responsiveness, all of Google’s AI calling features on Pixel phones run completely on-device. That includes Call Screen, but also Hold for Me, Direct My Call, Call Notes, AI-enhanced Audio, Scam Detection, and Live Translate. You can rest easy knowing the AI features in the Pixel phone app are safe to use, as they all run securely on-device, with many using Gemini Nano models.
Recording summaries
Quickly and privately summarize voice memos
Google’s Recorder app has been an attractive Pixel exclusive since it launched in 2019. The advanced voice memo app has the capability to transcribe recordings in real time, and the Recorder app has also used on-device processing since day one. That means your recordings are securely transcribed on your Pixel with zero data leaving your smartphone, and it also enables offline transcription. This was true long before AI became a tech buzzword and Gemini became a key brand, and it’s still true today.
Now, newer Pixel phones use the upgraded Gemini Nano with Multimodality model for real-time transcription, smarter AI summaries, and improved context understanding. This model is “more capable, accurate, and scalable,” according to Google, allowing the Pixel Recorder app to summarize longer transcripts. Crucially, it’s all happening on-device, enabling the speed of real-time transcription.
Now Playing
Like Shazam, but always running and always secure
Now Playing is one of the coolest Google Pixel features ever to releasebut it’s also one of the scariest to newcomers. When the Now Playing tool is enabled, your Pixel phone is always listening. If your smartphone hears a song, it’ll try to identify it, showing the name of the song and the artist on your lock screen. Additionally, it’ll save a list of recently-heard songs in the Now Playing menu of the Settings app on your Pixel.
So, if you hear a song you like while shopping or at a party and want to know the name, you don’t have to pull out Shazam or awkwardly ask Gemini — your Pixel will automatically find the track title for you, and you can check it at your own pace. The cool thing about Now Playing is that it uses a music library stored on-device to identify songs playing in your surroundings. It’ll need to download a song library over Wi-Fi the first time Now Playing is enabled, but after that, the feature works completely offline. Even though it’s always listening, song identification is handled securely on your Pixel.
Now Playing also includes an Enhance Now Playing toggle that is turned off by default. This option sends a “short, digital audio fingerprint” to Google to help identify tracks that can’t be named by the on-device database. Keep this toggle off if you want the privacy of on-device processing.
Pixel Screenshots
An AI-powered inventory of your screen captures
Pixel Screenshots is the dedicated app for screenshot management you’ve always wanted. Actually, it’s probably better than the one you’ve wanted, because it leverages on-device Gemini Nano models to process and organize your screenshots. This makes your screen captures searchable with natural-language requests. Gemini can extract and identify addresses, names, and other data points, making them available when you need them. For example, my Pixel Screenshots app automatically sorts my screenshots into Social media, Flight infoand Eventsand I can easily filter images as a result.
Pixel Screenshots can also give your screenshots a title, generate a summary of each screenshot, and offer suggestions about what to do next. For instance, if the app detects event information, it might suggest adding a Google Calendar event. Users can also create Collections in the Pixel Screenshots app to manually manage their screenshots. With the Recall feature, the app answers your questions in natural language using information stored in screenshots.
The app works quickly and securely on-device using Gemini Nano, meaning it’s available offline too.
Google’s on-device AI is powered by Gemini Nano
Some high-powered AI features, like Gemini Live, require cloud processing because their performance demands would overwhelm a smartphone chip. However, Google makes tools like Call Screen, Recorder summaries, Now Playing, and Pixel Screenshots available using a specialized AI model called Gemini Nano. It’s a small and efficient model that runs on the Google Pixel 8a and newer, enabling new features and allowing old ones to work offline for the first time.
Gemini Nano keeps getting better over time, with the latest version (available for the Pixel 9 series and later) adding multimodal support. With these features, you can use the best of Google AI without sending your data to cloud servers.