Feed: Engadget is a web magazine with obsessive daily coverage of everything new in gadgets and consumer electronics
Feed Link: https://www.engadget.com/
---

Title: Google I/O 2025 recap: AI updates, Android XR, Google Beam and
everything else announced at the annual keynote

Date: Tue, 20 May 2025 20:25:24 +0000
Link: https://www.engadget.com/ai/google-io-2025-re...

Today is one of the most important days on the tech calendar as Google kicked
off its I/O developer event with its annual keynote. As ever, the company had
many updates for a wide range of products to talk about.

The bulk of the Android news was revealed last week, during a special edition
of The Android Show. However, Tuesday's keynote still included a ton of stuff
including, of course, a pile of AI-related news. We covered the event in real-
time in our live blog, which includes expert commentary (and even some
jokes!) from our team.

If you're on the hunt for a breakdown of everything Google announced at the
I/O keynote, though, look no further. Here are all the juicy details worth
knowing about:

AI Mode chatbot is coming to Search for all US users

Quelle surprise, Google is continuing to shove more generative AI features
into its core products. AI Mode, which is what the company is calling a new
chatbot, will soon be live in Search for all US users.

AI Mode is in a separate tab and it's designed to handle more complex queries
than people have historically used Search for. You might use it to compare
different fitness trackers or find the most affordable tickets for an
upcoming event. AI Mode will soon be able to whip up custom charts and
graphics related to your specific queries too. It can also handle follow-up
questions.

The chatbot now runs on Gemini 2.5. Google plans to bring some of its
features into the core Search experience by injecting them into AI Overviews.
Labs users will be the first to get access to the new features before Google
rolls them out more broadly.

Meanwhile, AI Mode is powering some new shopping features. You'll soon be
able to upload a single picture of yourself to see what a piece of clothing
might look like on a virtual version of you.

Also, similar to the way in which Google Flights keeps an eye out for price
drops, Google will be able to let you know when an item you want (in its
specific size and color) is on sale for a price you're willing to pay. It can
even complete the purchase on your behalf if you want.

1.5 billion people see AI Overviews each month

AI Overviews, the Gemini-powered summaries that appear at the top of search
results and have been buggy to say the least, are seen by more than 1.5
billion folks every month, according to Google. The "overwhelming majority"
of people interact with these in a meaningful way, the company said ΓÇö this
could mean clicking on something in an overview or keeping it on their screen
for a while (presumably to read through it).

Still, not everyone likes the AI Overviews and would rather just have a list
of links to the information they're looking for. You know, like Search used
to be. As it happens, there are some easy ways to declutter the results.

Another look at Google's universal AI assistant

We got our first peek at Project Astra, Google's vision for a universal AI
assistant, at I/O last year and the company provided more details this time
around. A demo showed Astra carrying out a number of actions to help fix a
mountain bike, including diving into your emails to find out the bike's
specs, researching information on the web and calling a local shop to ask
about a replacement part.

It already feels like a culmination of Google's work in the AI assistant and
agent space, though elements of Astra (such as granting it access to Gmail)
might feel too intrusive for some. In any case, Google aims to transform
Gemini into a universal AI assistant that can handle everyday tasks. The
Astra demo is our clearest look yet at what that might look like in action.

AI updates

Gemini 2.5 is here with (according to Google) improved functionality,
upgraded security and transparency, extra control and better cost efficiency.
Gemini 2.5 Pro is bolstered by a new enhanced reasoning mode called Deep
Think. The model can do things like turn a grid of photos into a 3D sphere of
pictures, then add narration for each image. Gemini 2.5's text-to-speech
feature can also change up languages on the fly. There's much more to it than
that, of course, and we've got more details in our Gemini 2.5 story.

You know those smart replies in Gmail that let you quickly respond to an
email with an acknowledgement? Google is now going to offer personalized
versions of those so that they better match your writing style. For this to
work, Gemini looks at your emails and Drive documents. Gemini will need your
permission before it plunders your personal information. Subscribers will be
able to use this feature in Gmail starting this summer.

Google Meet is getting a real-time translation option, which should come in
very useful for some folks. A demo showed Meet being able to match the
speaker's tone and cadence while translating from Spanish to English.

Subscribers on the Google AI Pro and Ultra (more on that momentarily) plans
will be able to try out real-time translations between Spanish and English in
beta starting this week. This feature will soon be available for other
languages.

Google

Gemini Live, a tool Google brought to Pixel phones last month, is coming to
all compatible Android and iOS devices in the Gemini app (which already has
more than 400 million monthly active users). This allows you to ask Gemini
questions about screenshots, as well as live video that your phone's camera
is capturing. Google is rolling out Gemini Live to the Gemini iOS and Android
app starting today.

Google Search Live is a similar-sounding feature. You'll be able to have a
"conversation" with Search about what your phone's camera can see. This will
be accessible through Google Lens and AI Mode.

A new filmmaking app called Flow, which builds on VideoFX, includes features
such as camera movement and perspective controls; options to edit and extend
existing shots; and a way to fold AI video content generated with Google's
Veo model into projects. Flow is available to Google AI Pro and Ultra
subscribers in the US starting today. Google will expand availability to
other markets soon.

Speaking of Veo, that's getting an update. The latest version, Veo 3, is the
first iteration that can generate videos with sound (it probably can't add
any soul or actual meaning to the footage, though). The company also suggests
that its Imagen 4 model is better at generating photorealistic images and
handling fine details like fabrics and fur than earlier versions.

Handily, Google has a tool it designed to help you determine if a piece of
content was generated using its AI tools. It's called SynthID Detector ΓÇö
naturally, it's named after the tool that applies digital watermarks to AI-
generated material.

According to Google, SynthID Detector can scan an image, piece of audio,
video or text for the SynthID watermark and let you know which parts are
likely to have a watermark. Early testers will be able to to try this out
starting today. Google has opened up a waitlist for researchers and media
professionals. (Gen AI companies should offer educators a version of this
tech ASAP.)

The new AI Ultra plan costs $250 per month Google

To get access to all of its AI features, Google wants you to pay 250 American
dollars every month for its new AI Ultra plan. There's really no other way to
react to this other than "LOL. LMAO." I rarely use either of those acronyms,
which highlights just how absurd this is. What are we even doing here? That's
obscenely expensive.

Anyway, this plan includes early access to the company's latest tools and
unlimited use of features that are costly for Google to run, such as Deep
Research. It comes with 30TB of storage across Google Photos, Drive and
Gmail. You'll get YouTube Premium as well ΓÇö arguably the Google product
that's most worth paying for.

Google is offering new subscribers 50 percent off an AI Ultra subscription
for the first three months. Woohoo. In addition, the AI Premium plan is now
known as Google AI Pro.

A second Android XR device has been announced

As promised during last week's edition of The Android Show, Google offered
another look at Android XR. This is the platform that the company is working
on in the hope of doing for augmented reality, mixed reality and virtual
reality what Android did for smartphones. After the company's previous
efforts in those spaces, it's now playing catchup to the likes of Meta and
Apple.

The initial Android XR demo at I/O didn't offer much to get too excited about
for now. It showed off features like a mini Google Map that you can access on
a built-in display and a way to view 360-degree immersive videos. We're still
waiting for actual hardware that can run this stuff.

Xreal

As it happens, Google revealed the second Android XR device. Xreal is working
on Project Aura, a pair of tethered smart glasses. We'll have to wait a bit
longer for more details on Google's own Android XR headset, which it's
collaborating with Samsung on. That's slated to arrive later this year.

A second demo of Android XR was much more interesting. Google showed off a
live translation feature for Android XR with a smart glasses prototype that
the company built with Samsung. That seems genuinely useful, as do many of
the accessibility-minded applications of AI. Gentle Monster and Warby Parker
are making smart glasses with Android XR too. Just don't call it Google Glass
(or do, I'm not your dad).

Chrome's password manager is getting an upgrade

Google is giving the Chrome password manager a very useful weapon against
hackers. It will be able to automatically change passwords on accounts that
have been compromised in data breaches. So if a website, app or company is
infiltrated, user data is leaked and Google detects the breach, the password
manager will let you generate a new password and update a compatible account
with a single click.

The main sticking point here is that it only works with websites that are
participating in the program. Google's working with developers to add support
for this feature. Still, making it easier for people to lock down their
accounts is a definite plus. (And you should absolutely be using a password
manager if you aren't already.)

On the subject of Chrome, Google is stuffing Gemini into the browser as well.
The AI assistant will be able to answer questions about the tabs you have
open. You'll be able to access it from the taskbar and a new menu at the top
of the browser window.

Beam is the new name of Google's 3D video conferencing booths

It's been a few years since we first heard about Project Starline, a 3D video
conferencing project. We tried this tech out at I/O 2023 and found it to be
an enjoyable experience.

Now, Google is starting to sell this tech, but only to enterprise customers
(i.e. big companies) for now. It's got a new name for all of this too: Google
Beam. And it's probably not going to be cheap. HP will reveal more details in
a few weeks.

This article originally appeared on Engadget at
https://www.engadget.com/ai/google-io-2025-re...
beam-and-everything-else-announced-at-the-annual-keynote-
175900229.html?src=rss

---
VRSS v2.1.180528