Google Photos

• 1. Introduction

  ›

  Google Photos began in May 2015, when Google split it off from Google+, the company's former social network. Free of that network, the service posed a strange question: could software actually look at a photograph and know what was in it?

  Reviewers who saw the standalone service that same month called it a serious rival to every other cloud photo product on the market. Over the next several years, hundreds of millions of people would trust it with their photos.

  The same recognition technology that sorted photos by face and place would also make a very public mistake. And nearly a decade later, that same facial recognition would surface again, this time far from anyone's photo album. How did a spinoff from a failed social network become essential to so many people, and where did its machine vision go wrong?

• 2. Splitting From Google+

  ›

  Facebook was the network Google had set out to beat when it built Google+ years earlier, hoping to win the same audience for social sharing. Google+ never caught on the way Facebook did, either for socializing or for sharing photos. Its photo storage and organizing tools actually surpassed Facebook's own in power, but Google+ never built the user base to make full use of them. Leaving that social network behind changed what Photos was: it moved from a sharing platform into a private library platform.

  In December 2015, Google added a shared albums feature, letting users pool photos and videos into a single album for others to see. Google described the recipient's experience directly: they "can join to add their own photos and videos, and also get notifications when new pics are added." Users could also pull photos and videos out of a shared album and save them into their own private collection. Unlike the native Photos app on iOS, Google Photos allowed full resolution sharing across both Android and iOS, and between the two platforms.

  On the 12th of February 2016, Google announced that its Picasa desktop application would be discontinued on the 15th of March that year. Picasa Web Albums, the companion web service, was set to close on the 1st of May 2016. Google explained the decision plainly: it wanted to focus its efforts "entirely on a single photo service," the cross-platform, web-based Google Photos. That single-service strategy would soon expand far beyond Picasa, into automatic albums and a search bar built to recognize faces and places on its own.

• 3. Teaching A Camera To See

  ›

  Three categories organize every search in Google Photos: People, Places, and Things. The service's computer vision groups similar faces together, whether the face belongs to a human or a pet. That grouping only works in certain countries, because of privacy laws there. It recognizes geographic landmarks too; a photo containing the Eiffel Tower, for instance, gets tagged by place automatically. The Things category sorts by subject matter: birthdays, buildings, cats, concerts, food, graduations, posters, and screenshots among them. Users can manually remove a mistaken tag if the system gets a category wrong.

  The service can identify a scene as specific as "a snowstorm in Toronto" without anyone labeling it by hand. It can also flag scenes as ordinary as a beach or a skyline. Google Lens is built directly into the app, extending that same recognition to whatever a user points a camera at within Photos. People who receive a shared album can view the pictures in a web gallery without installing the app at all. Editing controls use finger swipes across the screen rather than sliders. The app can also generate a web link that both Google Photos users and non-users can open.

  A heat map showing where a person's photos were taken was added to the service in 2020. The system can also track a person's face as it ages across years of photos, matching a childhood picture to an adult one. Shared images can be posted onward to other social networks, including Google+, Facebook, and Twitter, alongside other services entirely. None of this recognition happened all at once; Google kept adding to it, feature by feature, for years after launch.

• 4. A Steady Stream Of Updates

  ›

  In June 2016, Google taught Photos to build albums on its own after a trip or an event. Each album came with a map of the traveled route and pins marking exact locations. Users could add their own text captions describing individual photos inside those automatic albums. By October, Google added several updates at once. Memories now surfaced people identified in a user's recent photos. The app also occasionally highlighted a batch of images when someone had recently photographed one subject repeatedly. It began animating short clips pulled from longer videos, not just still photos, and started detecting sideways or upside down images to suggest a fix. Google described all of this as the work of machine learning, requiring no input from the user.

  That November, Google released a separate app called PhotoScan, built for scanning printed photographs into the service on iOS and Android. Users centered their camera over four dots overlaying the printed photo. The software then stitched the shots into one high resolution digital image with the fewest possible defects. Later that same month, a new "Deep blue" slider let users adjust the color and saturation of skies in a photo. It did this without degrading quality or shifting the colors of anything else in the frame.

  February 2017 brought a redesigned Albums tab on Android, split into three sections. One section covered the phone's camera roll, with sorting by people or location. Another covered photos taken inside other apps, and a third covered the actual photo albums. In March, an automatic white balance feature reached the Android app and website first, with the iOS app following later. That same month, Google introduced lightweight preview uploads. A lower quality image appeared instantly on a slow cellular connection. A full quality version swapped in later once the phone reached Wi-Fi. The same low-to-high quality swap applied to sharing photos with other people. April brought video stabilization, which works by creating a duplicate of the video so the original clip is never overwritten.

  May 2017 brought the service's biggest batch of updates yet. Suggested Sharing reminded users to share photos after the fact and recommended recipients using facial recognition. Shared Libraries let two people maintain one central repository covering all of their photos or just specific categories of them. Photo Books turned digital collections into physical softcover or hardcover albums, with Photos auto suggesting which pictures to include based on face, location, trip, or other distinctions. Near the end of the month, an Archive feature let users hide photos from the main timeline without deleting them, while archived photos still turned up in albums and in search. The new sharing features from that May announcement began reaching users the following month, in June.

  In December 2018, Google doubled how many photos a private Live Album could hold, from 10,000 to 20,000, matching the capacity available to shared albums. By September 2019, a feature called "Memories" arrived, echoing the Stories format on Instagram and Facebook by resurfacing past photos for a nostalgic effect. On the 25th of June 2020, Google gave the mobile and web apps a major redesign, paired with a new, simplified logo. None of these updates changed what it cost to keep the free storage that made all of it possible.

• 5. Free Storage Has A Price

  ›

  Google Photos shares its 15 gigabytes of free storage with Gmail and Google Drive, one pool split across all three services. Uploads count toward that pool whether a user picks the original quality setting or a compressed one. The compressed tier covers photos up to 16 megapixels and video up to 1080p. Anything compressed and uploaded before the 1st of June 2021 does not count against the limit at all. Neither does anything uploaded through a Pixel phone released before that date. Anyone who needs more room can pay for it through a Google One subscription.

  Three storage settings exist inside Google Photos today: "Storage saver," formerly called "High quality"; "Original quality"; and "Express quality," which is not available everywhere. Storage saver caps photos at 16 megapixels and video at 1080p, the resolutions an average smartphone could manage back in 2015. Original quality keeps every photo and video exactly as it was captured, at full resolution. Express quality compresses further still, down to 3 megapixel photos and 480p video.

  The first three generations of Google Pixel phones came with unlimited free storage at Original quality, though the terms shifted with each generation. The original Pixel carried no cutoff date at all for that unlimited offer. The Pixel 2 kept unlimited Original quality storage only for photos and videos captured before the 16th of January 2021. The Pixel 3 held the same deal but with a cutoff of the 31st of January 2022. Photos and videos taken after each phone's cutoff date were uploaded at Storage saver instead. Starting with the Pixel 3a, Google dropped unlimited Original quality storage entirely; the Pixel 4, Pixel 4a, Pixel 4a 5G, and Pixel 5 instead came with a three month trial of the 100 gigabyte Google One plan for new subscribers.

  In November 2020, Google announced it would end free unlimited storage for photos uploaded at High quality or Express quality. The change would take effect on the 1st of June 2021, driven by rising demand for storage. On that same June date, "High quality" was formally renamed "Storage saver." The change was part of a broader push to lean less on advertising revenue and more on paid subscriptions. Photos already stored were unaffected. Only new uploads began counting against the shared quota across Drive, Gmail, and Photos. Pixel owners up through the Pixel 5 stayed exempt from the change entirely.

  GPS location data is one of the details users lose when exporting photos out of Google Photos, along with full original quality. Even Google's own Takeout tool, built for exporting data, leaves some of that original information behind. For anyone who has come to depend on Photos features that need that missing data, leaving the service can mean losing access to those features. None of these storage changes slowed how many people kept signing up.

• 6. Explosive Growth

  ›

  By October 2015, five months after launch, Google Photos had reached 100 million users, who had uploaded 3.72 petabytes of photos and videos between them. One year after release, in May 2016, that number had roughly doubled to over 200 million monthly active users. Google shared other figures at that milestone. At least 13.7 petabytes of photos and videos had been uploaded by then. The service had applied 2 trillion labels to images, 24 billion of them for selfies alone. Users had also generated 1.6 billion animations, collages, and effects from their own content.

  By May 2017, two years after launch, Google Photos passed 500 million users. Those users were now uploading over 1.2 billion photos to the service every single day. In 2019, four years after its initial release, Google Photos passed the 1 billion user mark.

  By November 2020, Google reported that more than 4 trillion photos were stored in the service altogether. Roughly 28 billion new photos and videos were being uploaded to the service every week by that point. Numbers that size do not explain why so many people trusted the service with their photos in the first place.

• 7. What The Critics Saw

  ›

  Walt Mossberg of Recode reviewed the newly independent service and called it the best in cloud photo storage. He measured it against competitors like Amazon Drive, Apple's iCloud, Dropbox, and Microsoft's OneDrive. He described the release as "liberation day" for photo features that had been "effectively hidden" inside what he called the "widely ignored" Google+ social network. Mossberg was especially taken with the search function; typing "Massachusetts" "instantly brought up loads of photos of subjects" tied to that place. He called the face grouping feature "remarkably accurate." He was even more impressed by subject based grouping. He was once surprised that a search for "boats" turned up both Cape Cod fishing boats and Venetian gondolas. He did note errors too, including one professional photograph that the system mistakenly registered as a screenshot.

  Jacob Kastrenakes of The Verge argued the release turned Google into a major competitor in the photo storage market. He wrote that its pricing made paying for photo storage feel outdated. He called the underlying service "always excellent" and welcomed the fact that people could use it "without signing up for a new social network." Kastrenakes also compared the new image analysis to technology Flickr had unveiled that same month.

  Sarah Mitroff and Lynn La of CNET singled out the phone and tablet apps as particularly strong. They called the design more streamlined than Yahoo's Flickr and better organized than Apple's iCloud photo service. Josh Lowensohn, also of The Verge, described the underlying strategy as putting all of a user's data on Google's servers so it could be reached from anywhere. Lowensohn praised the service's speed and intelligence, particularly its knack for sorting unorganized photos, along with its loading times, search speeds, and simple editing tools.

  PC Magazine's John C. Dvorak took a different view, focused on privacy rather than performance. He worried about Google's underlying motivation for building the service, its relationships with governments, and any future law that might force Google to hand over a user's entire photo history. Dvorak compared that scenario to inviting someone to "scrounge through your underwear drawer." He also criticized the service's sync functions, preferring folders of images to what he called an unsorted "flat database." He pointed to poor choices in which photos got animated as well. Citing Google's abrupt cancellation of Google Reader, Dvorak doubted the service's staying power. He suggested people use a portable hard drive instead, calling it "safer and cheap." Dvorak's worry about governments compelling Google to hand over photo histories would resurface in a very different form years later.

• 8. When The Algorithm Failed

  ›

  In June 2015, Google Photos's new image labeling feature identified a black software developer, Jacky Alcine, and a friend as "gorillas." The underlying dataset used to train the system had been built mostly from images of white people. Very few images of black people were included, a problem researchers call "sample size disparity." Google did not fix the bias in the algorithm itself. Instead, it applied a patch that stopped the system from ever labeling anything as a "gorilla." Eight years later, Google Photos was still unable to identify an actual gorilla in a photo.

  In March 2024, The New York Times reported that Google Photos was being used inside a facial recognition program. The program was run by Unit 8200, a surveillance unit of the Israeli Defense Forces. According to intelligence officers who spoke to the Times, the unit uploaded databases of known faces into Google Photos. It then used the service's search functions to identify individuals amid the war in Gaza. A Google spokesman responded that the service is free to use and "does not provide identities for unknown people in photographs." That spokesman's assurance, that the service does not name strangers in a photograph, now sits beside a search bar built to recognize one particular face among a vast and growing library of them.