OpenAI just announced that it recently conducted a small-scale preview of a new tool called Voice Engine. This is a voice cloning technology that can mimic any speaker by analyzing a 15-second audio sample. The company says it generates “natural-sounding speech” with “emotive and realistic voices.”

The technology is based on the company’s pre-existing text-to-speech API and it has been in the works since 2022. OpenAI has already been using a version of the toolset to power the preset voices available in the current text-to-speech API and the Read Aloud feature. There are a bunch of samples on the company’s official blog and they sound eerily close to the real thing. I encourage you to give them a listen and imagine the possibilities, both good and bad.

OpenAI says they see this technology being useful for reading assistance, language translation and helping those who suffer from sudden or degenerative speech conditions. The company brought up a Brown University pilot program that helped a patient with speech impairment issues by creating a Voice Engine clone pulled from audio recorded for a school project.

Despite the potential benefits, bad actors would certainly abuse this technology to engage in some serious deepfake tomfoolery, which is already a problem. With this in mind, Voice Engine isn’t quite ready for prime time, as there are serious privacy concerns that must be met before a full rollout.

OpenAI acknowledges that this tech has “serious risks, which are especially top of mind in an election year.” The company says its incorporating feedback from “US and international partners from across government, media, entertainment, education, civil society and beyond” to ensure the product launches with a minimal amount of risk. All preview testers agreed to OpenAI’s usage policies, which ban the impersonation of another individual without consent or legal right.

Additionally, anybody using the tech will have to disclose to their audience that the voices are AI-generated. OpenAI implemented safety measures, like watermarking to trace the origin of any audio and “proactive monitoring” of how the system is being used. When the product officially rolls out there will be a “no-go voice list” that detects and prevents AI-generated speakers that are too similar to prominent figures.

As for when that rollout will occur, OpenAI remains tight-lipped. TechCrunch uncovered some potential pricing data and it looks like it will undercut competitors in the space like ElevenLabs. Voice Engine could cost $15 per one million characters, which works out to around 162,500 words. This is about the length of Stephen King’s The Shining. It certainly sounds like a budget-friendly way to get an audiobook done. The marketing materials also make reference to an “HD” version that costs twice as much, but the company hasn’t detailed how that will work.

OpenAI has been making big moves this week. It just announced another partnership with its bestie Microsoft to build an AI-based supercomputer called “Stargate.” The project will reportedly cost a whopping $100 billion, according to The Information.

X is working on features that will allow admins of “Communities,” the platform’s tool for subreddit-like groups, to designate the spaces as containing “adult content.” The change was confirmed by an engineer at X amid reports that the Elon Musk-owned company was working on enabling NSFW groups.

In a post on X, engineer Dong Wook Chung noted that “soon” NSFW content would be automatically filtered in the app’s Communities feature. “Admins can now set 'Adult content' in Settings to avoid auto-filtering of the content,” Chung said.

As Bloomberg reported, researchers had previously spotted clues that X planned to enable settings for “adult-sensitive” content. X permits users to share nudity and other “graphic” content, but doesn’t allow it to appear in certain parts of the app, like profile photos and cover images for Communities.

X’s Communities feature predates Musk’s takeover of the company. Twitter began experimenting with the idea in 2021, saying it would provide “a more intimate space for conversations” on the platform. Though Twitter never publicly discussed enabling NSFW features for Communities, the app allowed adult content, unlike most of its social media peers. The company reportedly looked into creating an OnlyFans competitor with its creator subscription product in 2022. The plan was eventually scrapped, according to the Platformer newsletter, due to concerns it would “worsen” the company’s problems with illegal child exploitation content.

It’s not clear if X’s current leadership has addressed those concerns. In a separate post, Chung, the X engineer, stated that the new filtering settings “is about making Communities safer for everyone by automatically filtering out” adult content. “Only users who have specified their age will be able to search Communities with NSFW content.” 

X didn’t immediately respond to a request for comment.

In 2014, WIRED asked me to write a few lines about my most-used app as part of an internship application. I wrote about WhatsApp because it was a no-brainer. I was an international student from India, and it was my lifeline to my family and to my girlfriend, now my wife, who lived on the other side of the world. “This cross-platform messenger gets all the credit for my long-distance relationship of two years, which is still going strong,” I wrote in my application. “Skype is great, Google+ Hangouts are the best thing to have happened since Gmail but nothing says ‘I love you’ like a WhatsApp text message.”

A few months into that internship, Facebook announced it was buying WhatsApp for a staggering $19 billion. In WIRED’s newsroom, there were audible gasps at this seemingly minor player's price tag. American journalists weren’t exactly unfamiliar with WhatsApp. Much of the country was still locked in a battle between green and blue bubbles, even as the rest of the world had switched to an app created by two former Yahoo! engineers in WIRED’s Mountain View backyard.

Text messaging was one of the few things you could do on WhatsApp in 2014. There were no emoji you could react with, no high-definition videos you could send, no GIFs or stickers, no read receipts until the end of that year and certainly no voice or video calling. And yet, more than 500 million people around the world were hooked, reveling in the freedom of using nascent cellular data to swap unlimited messages with friends and family instead of paying mobile carriers per text.

WhatsApp’s founders, Jan Koum and Brian Acton, launched the app in 2009 simply to display status messages next to people’s names in a phone’s contact book. But after Apple introduced push notifications on the iPhone later that year, it evolved into a full-blown messaging service. Now, 15 years later, WhatsApp has become a lot more — an integral part of the propaganda machinery of political parties in India and Brazil, a way for millions of businesses to reach customers, a way to send money to people and merchants, a distribution platform for publications, brands and influencers, a video conferencing system and a private social network for older adults. And it is still a great way for long-distance lovers to stay connected.

“WhatsApp is kind of like a media platform and kind of like a messaging platform, but it’s also not quite those things,” Surya Mattu, a researcher at Princeton who runs the university’s Digital Witness Lab, which studies how information flows through WhatsApp, told Engadget. “It has the scale of a social media platform, but it doesn’t have the traditional problems of one because there are no recommendations and no social graph.”

Indeed, WhatsApp’s scale dwarfs nearly every social network and messaging app out there. In 2020, WhatsApp announced it had more than two billion users around the world. It’s bigger than iMessage (1.3 billion users), TikTok (1 billion), Telegram (800 million), Snap (400 million) and Signal (40 million.) It stands head and shoulders above fellow Meta platform Instagram, which captures around 1.4 billion users. The only thing bigger than WhatsApp is Facebook itself, with more than three billion users .

WhatsApp has become the world’s default communications platform. Ten years after it was acquired, its growth shows no sign of stopping. Even in the US, it is finally beginning to break through the green and blue bubble battles and is reportedly one of Meta’s fastest-growing services. As Meta CEO Mark Zuckerberg told the New York Times last year, WhatsApp is the “next chapter” for the company.

Will Cathcart, a longtime Meta executive, who took over WhatsApp in 2019 after its original founders departed the company, credits WhatsApp’s early global growth to it being free (or nearly free — at one point, WhatsApp charged people $1 a year), running on almost any phone, including the world’s millions of low-end Android devices, reliably delivering messages even in large swathes of the planet with suboptimal network conditions and, most importantly, was dead simple, free of the bells and whistles that bloat most other messaging apps. In 2013, a year before Facebook acquired it, WhatsApp added the ability to send short audio messages.

“That was really powerful,” Cathcart told Engadget, “People who don’t have high rates of literacy or someone new to the internet could spin up WhatsApp, use it for the first time and understand it.”

In 2016, WhatsApp added end-to-end encryption, something Cathcart said was a huge selling point. The feature made WhatsApp a black box, hiding the contents of messages from everyone — even WhatsApp — except the sender and the receiver. The same year, WhatsApp announced that one billion people were using the service every month.

That explosive growth came with a huge flip side: As hundreds of millions of people in heavily populated regions, like Brazil and India, came online for the first time, thanks to inexpensive smartphone and data prices, WhatsApp became a conduit for hoaxes and misinformation to flow freely. In India, currently WhatsApp’s largest market with more than 700 million users, the app overflowed with propaganda and disinformation against opposition political parties, cheerleading Narendra Modi, the country’s nationalist Prime Minister accused of destroying its secular fabric.

Then people started dying. In 2017 and 2018, frenzied mobs in remote parts of the country high on baseless rumors about child abductors forwarded through WhatsApp, lynched nearly two dozen people in 13 separate incidents. In response to the crisis, WhatsApp swung into action. Among other things, it made significant product changes, such as clearly labeling forwarded messages — the primary way misformation spread across the service — as well as severely restricting the number of people and groups users could forward content to at the same time.

In Brazil, the app is widely seen as a key tool in the country’s former President Jair Bolsonaro’s 2018 win. Bolsonaro, a far-right strongman, was accused of getting his supporters to circumvent WhatsApp’s spam controls to run elaborate misinformation campaigns, blasting thousands of WhatsApp messages attacking his opponent, Fernando Haddad.

Since these incidents, WhatsApp has established fact-checking partnerships with more than 50 fact-checking organizations globally (because WhatsApp is encrypted, fact-checkers depend on users reporting messages to their WhatsApp hotlines and respond with fact checks). It also made additional product changes, like letting users quickly Google a forwarded message to fact-check it within the app. “Over time, there might be more things we can do,” said Cathcart, including potentially using AI to help with WhatsApp’s fact-checking. “There’s a bunch of interesting things we could do there, I don’t think we’re done,” he said.

Recently, WhatsApp has rapidly added new features, such as the ability to share large files, messages that auto-destruct after they’re viewed, Instagram-like Stories (called Statuses) and larger group calls, among other things. But a brand new feature rolled out globally in fall 2023 called Channels points to WhatsApp’s ambitions to become more than a messaging app. WhatsApp described Channels, in a blog post announcing the launch, as “a one-way broadcast tool for admins to send text, photos, videos, stickers and polls.” They’re a bit like a Twitter feed from brands, publishers and people you choose to follow. It has a dedicated tab in WhatsApp, although interaction with content is limited to responding with emoji — no replies. There are currently thousands of Channels on WhatsApp and 250-plus have more than a million followers each, WhatsApp told Engadget. They include Puerto Rican rapper Bad Bunny (18.9 million followers), Narendra Modi (13.8 million followers), FC Barcelona (27.7 million followers) and the WWE (10.9 million followers). And even though it’s early days, Channels is fast becoming a way for publishers to distribute their content and build an audience.

“It took a year for us to grow to an audience of 35,000 on Telegram,” Rachel Banning-Lover, the head of social media and development at the Financial Times (155,000 followers) told Nieman Lab in November. “Comparatively, we [grew] a similar-sized following [on WhatsApp] in two weeks.”

WhatsApp’s success at consistently adding new functionality without succumbing to feature sprawl has allowed it to thrive, both with its core audience and also, more recently, with users in the US. According to data that analytics firm Data.ai shared with Engadget, WhatsApp had nearly 83 million users in the US in January 2024, compared to 80 million a year before. A couple of years ago, WhatsApp ran an advertising campaign in the US — its first in the country — where billboards and TV spots touted the app’s focus on privacy.

It’s a sentiment shared by Zuckerberg himself, who, in 2021, shared a “privacy-focused vision for social networking” on his Facebook page. “I believe the future of communication will increasingly shift to private, encrypted services where people can be confident that what they say to each other stays secure and their messages and content won’t stick around,” he wrote. “This is the future I hope we will help bring about.”

Meta has now begun using WhatsApp’s sheer scale to generate revenue, although it’s unclear so far how much money, if any, the app makes. “The business model we’re really excited about and one that we’ve been growing for a couple of years successfully is helping people talk to businesses on WhatsApp,” Cathcart said. “That’s a great experience.” Meta monetizes WhatsApp by charging large businesses to integrate the platform directly into existing systems they use to manage interactions with customers. And it integrates the whole system with Facebook, allowing businesses to place ads on Facebook that, when clicked, open directly to a WhatsApp chat with the business. These have become the fastest-growing ad format across Meta, the company told The New York Times.

A few years ago, a configuration change in Facebook’s internal network knocked multiple Facebook services, including WhatsApp, off the internet for more than six hours and ground the world to a halt.

“It’s like the equivalent of your phone and the phones of all of your loved ones being turned off without warning. [WhatsApp] essentially functions as an unregulated utility,” journalist Aura Bogado reportedly wrote on then-Twitter. In New Delhi and Brazil, gig workers were unable to reach customers and lost out on wages. In London, crypto trades stopped as traders were unable to communicate with clients. One firm claimed a drop of 15 percent. In Russia, oil markets were hit after traders were unable to get in touch with buyers in Europe and Asia placing orders.

Fifteen years after it was created, the messaging app runs the world.

To celebrate Engadget's 20th anniversary, we're taking a look back at the products and services that have changed the industry since March 2, 2004.

Elon Musk's answer to ChatGPT is getting an update to make it better at math, coding and more. Musk's xAI has launched Grok-1.5 to early testers with "improved capabilities and reasoning" and the ability to process longer contexts. The company claims it now stacks up against GPT-4, Gemini Pro 1.5 and Claude 3 Opus in several areas. 

Going by xAI's numbers, Grok-1.5 appears to be a large improvement over Grok-1. It shot up to 50.6 percent in the MATH benchmark, over double the previous score. It also climbed to 90 percent and 74.1 percent in GSM8K (math word problems) and HumanEval (coding), respectively, compared to 62.9 percent and 63.2 percent before. Those numbers are within shouting distance of Gemini Pro 1.5, GPT-4 and Claude 3 Opus — in fact, the HumanEval coding score beats all rivals except Claude 3 Opus.

xAI

It can also process long contexts of up to 128K tokens within its context window, meaning it can amalgamate data from more sources to understand a situation. "This allows Grok to have an increased memory capacity of up to 16 times the previous context length, enabling it to utilize information from substantially longer documents," the company said.

xAI didn't detail Grok's progress in other areas, though, where it still may be lagging (academic scores, multimodal and others). And Grok-1.5 may not keep its position for long. ChatGPT 5 is set to arrive sometime this summer, promising a feature set that "makes it feel like you are communicating with a person rather than a machine," according to OpenAI. 

Currently, Grok is only available for users of the Premium+ tier on X (formerly Twitter), though Elon Musk recently promised to open it up to X's regular Premium users. The company also recently open sourced its Grok chatbot, after Musk sued OpenAI and Sam Altman for allegedly abandoning its non-profit mission. 

Activision is reportedly in the midst of investigating a hacking campaign that's stealing login credentials from people playing its games. According to TechCrunch, bad actors have been successfully installing malware onto victims' computers and using their access to steal logins for their gaming accounts and even their crypto wallets. The video game publisher has apparently been helping victims remove the malware and regain control of their accounts, but it doesn't have enough information at the moment to say how the malware is spreading. 

TechCrunch's source said the malware "could be only affecting folks who have third-party tools installed," insinuating that people are getting it from non-Activision-developed software typically used with its games. Delaney Simmons, Activision's spokesperson, told the publication that the company is aware of "claims that some player credentials across the broader industry could be compromised from malware from downloading or using unauthorized software." He added that the company's servers "remain secure and uncompromised."

That's certainly a plausible theory, seeing as the hacking scheme appears to have been uncovered by someone known as Zeebler, who develops cheating software for Call of Duty. Zeebler told TechCrunch that he discovered the campaign when one of his customers had their account stolen for his software. Upon looking into it, he reportedly discovered a database containing stolen credentials. He also said that the malware is disguised to look like real software, but they were actually designed to steal the usernames and passwords victims type in. Zeebler is presumably talking about third-party tools like cheating software getting cloned to harvest people's logins, but phishing schemes that use Activision's official login design exist, as well. Bottom line is, people should be careful what they download and always double check if the login page they're typing in is the real deal. 

Snapchat has a new AI-powered perk for subscribers: Bitmoji versions of your pet. The feature, which is unfortunately not called “petmoji,” allows users to snap a photo of their four-legged friend to create a cartoon-like avatar to accompany their Bitmoji in the Snap Map.

Based on screenshots shared by the company, it seems users will be able to choose from a few different variations of the AI-generated images after sharing a photo of their pet. That’s considerably less customization than what you can do with your own human-inspired Bitmoji,though it should allow users to create something that looks similar to their IRL pet. (No word on if Snap could one day introduce branded pet accessories for animal avatars like they do for human Bitmoji.)

The addition is also the latest example of how Snap has embraced AI features in its subscription offering. Since debuting Snapchat+ in 2022, the company has used the premium service to experiment with generative AI features, including its MyAI assistant as well as camera-powered features like Dreams and AI-generated snaps. Snapchat+ has more than 7 million subscribers, the company announced in December.

Elsewhere, Snap added some updates for non-subscribers, too. The app is adding a new template feature to make it easier to edit clips, and new swipe-based gestures to send and edit snaps more quickly. Snapchat will also support longer video uploads for Stories and Spotlight. In-app captures can now be three minutes long, while the app will support uploads of up to five minutes.

Apple will finally launch new iPads in early May, according to Bloomberg’s Mark Gurman. Expected are a new iPad Pro with an OLED display and a faster iPad Air, including a 12.9-inch model for the first time in that lineup. The details of the upcoming iPad models have been consistent, circulating through the rumor mill since last year.

The new iPad Pro models will reportedly add OLED displays (offering deeper blacks and richer colors) and run on the new M3 chip, already found in several Macs. The new tablets are said to launch alongside a redesigned Apple Pencil and Magic Keyboard. Other than a white color option, the latter has remained unchanged since its arrival four years ago.

Meanwhile, the iPad Air will supposedly run on a new processor. Bloomberg didn’t specify which, but — considering the current model uses the M1, and Apple likely wants to reserve the M3 for the more expensive Pro — the M2 sounds like a safe bet. The 12.9-inch screen option would mark the first time the iPad Air line has offered a display larger than 10.9 inches. Although Apple will charge more for that model than the smaller sibling expected alongside it, that would be the cheapest way yet to get a supersized iPad screen.

Gurman said early this month that the new tablets would launch alongside the M3 MacBook Air, but the laptop arrived without any iPads in tow. He now reports that Apple’s release schedule was pushed back to finish working on the devices’ software and ironing out the kinks from the “complex new manufacturing techniques” they require.

A $3 iOS app now records higher-resolution spatial videos than Apple’s native camera app. Spatialify, available on the App Store, lets iPhone 15 Pro owners record 3D videos for Apple’s Vision Pro in either 1080p at 60fps or 4K at 30fps — with HDR. Apple’s native recording only supports 1080p / 30fps without HDR, so your immersive clips will be noticeably sharper using Spatialify than the camera app on the same phone. UploadVR first reported on the app update.

Spatialify launched earlier this year as a tool for converting Apple’s spatial videos (HEVC format) for playback on non-Apple VR headsets like the Meta Quest 3. But with Meta later adding native HEVC conversion to its headsets (the best-selling on the market), Spatialify’s superior recording could give the third-party app a new lease on life.

Engadget senior editor Devindra Hardawar confirmed that Spatialify produces files in 4K / 30fps when set accordingly. He also verified that Spatialify’s videos look much sharper on the Vision Pro than those shot in Apple’s camera app. HDR also makes the videos’ lighting look more realistic. Not bad for three bucks.

It’s somewhat surprising Apple is holding back its native camera app from exploiting the iPhone 15 Pro and Pro Max’s full hardware capabilities, but it isn’t unheard of. Halide, a popular iOS camera app, beat Apple to the punch with iOS photography features like shooting in RAW, manual controls and portrait photos for pets. Based on that history, I wouldn’t be shocked if Apple soon added similar advanced spatial recording to its camera, especially now that we know its current hardware has no problem with it.

Spatialify is available on the App Store for $3. It requires an iPhone 15 Pro or iPhone 15 Pro Max to capture spatial videos.

Amazon just launched an app that lets people sign up for its palm recognition service without having to head to an in-store kiosk. The Amazon One app uses a smartphone’s camera to take a photo of a palm print to set up an account. Once signed up, you can pay for stuff by using just your hand, ending the tyranny of having to carry a smartphone, cash or a burdensome plastic card.

The tech uses generative AI to analyze a palm's vein structure, turning the data into a “unique numerical, vector representation” which is recognized by scanning machines at retail locations. You’ll have to add a payment method within the app to get started and upload a photo of your ID for the purpose of age verification.

The app launches today for iOS and Android. Previously, you’d have to go to a physical location to sign up for Amazon One. Beyond payments, the tech is also used as an age verification tool and as a way to enter concerts and sporting events without having to bring along a ticket.

Once you hand over your palm-print to the completely benevolent Amazon corporation, you’ll have unfettered access to each and every Whole Foods grocery store throughout the country. Amazon, after all, owns Whole Foods. Amazon One payments are also accepted at some Panera Bread locations, in addition to certain airports, stadiums and convenience stores.

There are obvious privacy concerns here, as passwords can change but palms cannot. Amazon says that all uploaded palm images are “encrypted and sent to a secure Amazon One domain” in the Amazon Web Service cloud. The company also says the app “includes additional layers of spoof detection,” noting that it’s not possible to save or download palm images to the phone itself.

Google will bring Gemini, the company’s new large language model, to Pixel 8 smartphones after all. The phone will incorporate Gemini Nano, a version of the model built to run locally on personal devices. This follows a successful rollout to the Pixel 8 Pro late last year and the Samsung Galaxy S24 in January.

The Pixel 8 features the same proprietary Tensor G3 chip as the Pro, which was designed to speed up AI performance. So the overall experience should be similar with both gadgets. It’ll be coming in the next Pixel Feature Drop, but only as a developer preview for now. Google wants to collect feedback and make sure everything is running smoothly on the slightly lower-specced phone.

This is a fairly sudden change for Google. The company originally said that the Pixel 8 couldn't handle on-device Gemini because of "hardware limitations", despite having the same chip as the Pro model. The main difference between the two phones is the RAM allotment, which doesn't seem like a deal-breaker when it comes to running an on-device AI. It looks like Google also came around to that line of thinking. 

So what can you do with this thing? The company’s expanding two features that make use of the LLM, and both of these tools have been available for Pro users. The Recorder app will get an improved Gemini-powered summarize feature that works locally on the device. A similar tool already exists, but requires a network connection. Once Gemini Nano is on-board, the data will no longer have to get sent to a server. As one might surmise, this feature creates summaries of recorded conversations.

Gemini Nano will also power Gboard’s Smart Reply toolset. This software suggests responses to messages and is even aware of context in conversations. It started as an exclusive to WhatsApp but recently expanded to Line and KakaoTalk. Google promises “more messaging apps” will add support in the coming months. It’s pretty odd that Google Messages isn’t one of the early adopters of this platform. Gemini Nano is used to power on-device Magic Compose in Messages on the Pixel 8 Pro and Samsung S24, but Google hasn’t announced if this particular tool is coming to the standard Pixel 8.

This has been a big week for Gemini, as Apple is reportedly in talks to bring Google’s AI to iPhones. That’s like Coke and Pepsi suddenly opening up a soda shop together. However, Gemini recently came under fire for creating wildly inaccurate historical images. The image generation feature is still on pause as the company continues to iron out those kinks.