The Made by Google event showcased not only Google’s latest Pixel hardware but also launched numerous new AI features. I’m generally skeptical of current AI technologies, but after exploring the new software through various demos, I found myself increasingly intrigued. It appears Google, alongside Apple and Samsung, is focused on making AI-powered updates genuinely helpful, potentially simplifying our lives or adding a touch of fun. Due to time constraints, I couldn’t cover everything, so I’ve highlighted a few of my favorites to give you an idea of what to expect when the Pixel 10 series becomes available later this month. Notably, many of these advancements revolve around voice and calls – an area where Google has consistently excelled.
The Recorder app has long been a favorite of mine. It started with on-device transcription, which made obtaining quotes from interviews easy and secure. However, when Apple introduced multi-track recording to its Voice Memos app, I quickly switched. While the iOS recorder’s transcription isn't as accurate or readable, the ability to record a duet with myself was appealing. I played both Elphaba and Glinda, singing their parts from “For Good” into my iPhone. But when Google’s senior director of product management for Pixel software, Shenaz Zack (Shenaz Zack), told me the Pixel 10’s recorder app would add AI-generated music to your singing, I was taken aback.
I spent much of my youth ripping karaoke tracks from YouTube videos, searching for “minus one,” “backing tracks,” or “instrumentals only” on various platforms. My friends and I were aspiring performers, and a tool that would generate backing music for our voice tracks would have been a dream come true. It still feels that way. Zack demonstrated the process twice. On my first attempt, I sang a verse and part of the chorus of “Golden” from the Kpop Demon Hunters soundtrack. I felt self-conscious at the end, before Zack stopped the recording. As it recorded, the app displayed a tag indicating it recognized I was singing, and after selecting the recording, a chip appeared saying “Create and add music.” Tapping this brought up a panel titled “Choose a vibe to create music” with two sections: “Featured vibes” and “Your vibes.” The first section offered options like “Chill beats,” “Cozy,” “Dance party,” “Rainy day blues,” “Romantic,” and “Surprise me.”
On my second attempt, I quickly sang “Mary Had a Little Lamb,” and the app warned that the beat might not match a short recording. I chose “Dance Party” and waited a minute while Recorder analyzed the audio, identified the rhythm, locked onto the beat, and harmonized the track. I wasn’t sure what to expect, but those concerned about digital rights management have nothing to worry about. The music generated for “Golden” didn’t resemble the original, and while it made my voice less isolated and created a more complete track, I felt some adjustments were needed. As for “Mary Had a Little Lamb,” the result was as generic as expected for an AI-generated soundtrack to a nursery rhyme. However, the generated music was in the correct key and rhythm, and I'll need more time to experiment with settings. Notably, the music stopped as my singing stopped, so my giggling wasn’t included. Although this feature didn’t fulfill my fantasy, I see it as a fun and harmless use of AI. It likely won’t become a daily routine for most, although Zack mentioned a significant percentage of Recorder users already use it for singing. This update could add a touch of musical creativity.
I had concerns about the Voice Translate feature, designed to make callers sound like they're speaking in a different language. Google’s goal is to “break down language barriers during phone calls.” When I asked Zack why the company felt it was important to replicate the caller’s voice, she explained it was about personal connection. She mentioned her parents live in India and aren't fluent in English, making conversations with her children difficult. A robotic voice simply translating between them wouldn’t feel right.
Initially skeptical, I was swayed after a few demos. To be clear, the caller must use a Pixel phone for Voice Translate to work. After selecting Voice Translate from the Call Assist submenu, you choose a language. When the call connects, the system informs both parties that “Call is translated by Google AI in each speaker’s voice. Audio is not saved.” I tested this with a Google representative who spoke German, whom I’ll refer to as “Uncle Tim.” Each time he spoke, I heard a couple of seconds of his German voice before a chime and the translated version in English. The voice sounded like Uncle Tim, replicating pitch, rhythm, and expression. I also heard feedback when I spoke, hearing myself speaking German. It was strange, but it did sound like me.
A close friend lives in Germany and has endured my attempts to learn the language for over a decade. I immediately wanted to test Voice Translate on her to see if she’d believe I’d suddenly become fluent. The experience wasn’t perfect. The translations were sometimes inaccurate, and the generated voices weren’t complete replications, sounding more like novice dubbing artists. However, this was intentional, addressing concerns about impersonation. Zack explained that the implementation includes “ducking,” where the original speech is audible for a few seconds before fading below the dubbed voice. The AI voice translates the content but doesn’t invent what I say, allowing me to decide whether to curse a relative in their native tongue. There may still be bugs to work out, but I was amused by the various accents in the English version of Uncle Tim, which shifted from American to Australian.
This is powered by the Pixel 10’s Tensor G5 chip and processed on-device using a new codec and semantic understanding to interpret vocal expressions. I’m eager to call my friend in Frankfurt. At launch, Voice Translate will support translation to or from English with Spanish, German, Japanese, French, Hindi, Italian, Portuguese, Swedish, Russian, and Indonesian.
The Recorder app, translation, and expressive AI demonstrate Google’s long-standing expertise. The company has also been a pioneer in suggesting actions from emails and adding events to calendars. With the Pixel 10’s Magic Cue feature, Google is bringing this functionality to texts and calls. It can show shortcuts within the Messages app to answer questions about reservations or send photos. I’m particularly impressed by its ability to display flight reservation information during a call to an airline, eliminating the need to search for confirmation numbers. It’s a small convenience, but Google already offers a similar feature in its inbox. I’d like to see this expand to other appointments, quickly providing codes or identifying information during calls to plumbers, doctors, or insurance providers.
Google continues to improve its strengths in photography. The company was among the first to use algorithmic processing to dramatically improve low-light photos, and the Pixel 10 Pro continues this tradition. Pro Res Zoom produced surprisingly clear pictures of distant buildings during my demo at Google’s Manhattan office. The lines on the underside of a skyscraper, zoomed to 100x, were remarkably clear. Google clarified that Pro Res Zoom won’t work on people and that distant text may appear distorted. They’ve tuned it to minimize “hallucinations,” but they may still occur, especially with text. When Pro Res Zoom detects a person, it switches to Super Res Zoom to prevent inaccurate representations. Depending on the Pixel phone, Super Res Zoom delivers up to 20x or 30x zoom. People on a tower deck appeared pixelated compared to the building’s facade, but the effect wasn’t noticeable until zoomed in.
I was pleasantly surprised by the Camera Coach feature, which proposes framing options. I don’t think I’ll use it much, but it might help others learn photography. I was initially nonplussed by the new Photos feature that lets you use AI to edit pictures, but a brief demo changed my mind. Simply telling Gemini to “turn that red dress blue” or “get rid of the people in the background” was easy and effective. Google also tweaked the Guided Frame feature for visually impaired users, incorporating Gemini models to improve object recognition. Finally, it’s worth noting the support for C2PA content authenticity initiative. Google is building this into the Photos app, where metadata will indicate whether AI was used in a picture. The Pixel 10 phones will be the first to implement the new industry-standard Content Credentials (CR) within its native camera app, with companies like Adobe, Amazon, Google, Meta, Microsoft, and OpenAI participating in the initiative.
There were many other promising AI-related features demonstrated, like visual overlays in Gemini Live and the new Pixel Journal app. I didn’t spend much time with them, but they worked during my demos. The “take a message” feature, which transcribes voicemails, seems like a better way to be alerted to missed calls than a hidden section of the phone. I’m hopeful to see how these features evolve. I hope its competitors take note.