Could Generative AI Kill the Stock Photo?
Stock Images Could Face the Most Direct Disruption from Generative AI
Stock photo services like Shutterstock or Getty Images helped source the models used by tools like Midjourney and Stable Diffusion. But stock photos are not perfect—and if the two models can’t work together, one could disrupt the other.
If you’ve used an AI image generator for any period of time, you’ve most certainly seen more than a few watermarks for stock photo services get through.
It’s for that reason, among others, that a lawsuit between the stock photo service Getty Images and Stability AI, the makers of the text-to-image model Stable Diffusion, was filed back in February.
Getty Images isn’t necessarily against generative AI, as long as they have some control over the model and can use it to generate royalties for its photographers—it announced a collaboration with NVIDIA to this effect back in March.
But it’s nonetheless worth understanding why a stock photo service might react so strongly to generative AI as a concept. After all, let’s face it—this technology, as it develops, could replace the need for stock photos in many cases where they might have come in handy in the past.
And it comes down to understanding the roots of stock photography. Those roots can be found deep under the Earth’s surface.
From Iconic to Utilitarian: The Evolution of Stock Photos
In one of its final stories, BuzzFeed News completed a tour of the Bettmann Archive, a store of important photos organized by curator Otto Bettmann starting in the 1930s. It was a collection he built from a starting point of two utterly stuffed suitcases he brought to the United States after leaving a Nazi-controlled Germany in 1935. His archive—not the first, but easily the most famous of its kind—blossomed from there and ended up becoming one of the most important archives in the world. These photos, of which there are many millions, represent some of our most important images.
This is effectively our culture. Nearly every famous picture you can think of is included in this archive, and as a result, they are protected in a deep, cool setting—a Pennsylvania location more than 200 feet underground, roughly 50 miles from the nearest large city.
“The vault, which is part of a series of former limestone mines, is both temperature- and moisture-controlled. These careful conditions have virtually stopped the degradation of the photographic negatives,” the BuzzFeed piece notes.
These photos are largely in physical forms, in a location intended to keep their usability pristine for centuries to come. Odds are, however, that if you’ve seen these images, it’s largely in digital form. These kinds of photos, which often found use in newspapers and magazines, are often valuable for their artistic, cultural, or editorial implications, and their existence has created something of an arms race over the years.
Not all stock images need to be iconic or famous, however. Sometimes a photo of something plain or someone anonymous was enough—as it could be used as the starting point of a photo illustration, or could exemplify the type of person or thing the editors wished to include. (Pictures that show diversity are popular parts of stock photo libraries, for example.) In the past, these types of images were sold through catalogs produced by stock photo agencies which managed each individual photo.
But as the digital era came about, so did the methods of distribution. PhotoDisc, a company that started in the early ’90s, pioneered two key distribution techniques for the PC era—CD-ROM distributions (as their name suggests), and later, digital downloads.
(In case you’re in need of some stock images, many of the original PhotoDisc archives are available on the Internet Archive, including the ever-popular Objects Series, which is basically made up of a bunch of different things on a white background.)
At the time PhotoDisc started, most agencies were still focused on physical distribution, but that quickly changed in part to budding corporate interest at the very beginning of the internet era. The big shift happened thanks to investments by individuals with mega-rich reputations.
In 1994, Getty Oil heir Mark Getty began building an image archive with his business partner Jonathan Klein, with the explicit goal of repairing the “highly fragmented” state of stock photography, as a 1998 piece in The Independent put it.
This was the perfect time to build this kind of business. Mediums like books and magazines were having a visual renaissance thanks to the rise of desktop publishing, and designers needed images from which to illustrate stories. And Mark Getty, just 23 when his family’s oil business was sold off for $11 billion in the early ’90s, saw an opportunity in investing some of those earnings into the stock photo industry.
“We saw the market as a large but inefficient one,” Getty told The Independent. “We also saw technology as a catalyst for change in the industry … Most importantly, it did not seem to us that anyone else in the industry had yet grasped this opportunity.”
It turned out that Getty was wrong about that last part—almost immediately, Getty Images would find itself competing with, of all people, Bill Gates. Interactive Home Systems, a company founded by the Microsoft CEO in 1989, initially aimed to build art displays for homes. The company, in its pursuit of this initial goal, began to acquire a number of prominent image libraries, including the Bettmann Archive, which by this point had swelled to include much of United Press International’s legacy photo library—a total of 19 million images.
Eventually, Gates’ company, soon renamed Corbis Images, determined that licensing images to others was the pathway to success, and he had the high-end images that were worth buying.
But having pictures of Martin Luther King or Iwo Jima or the moon landing only goes so far in the world of stock photos—after all, mundane imagery likely makes up the bulk of stock photo sales. As exemplified by the rise of clip art, an artistic medium that ad agencies first used in the 1950s but which went mainstream starting in the 1980s and 1990s, people sometimes just need basic pictures to illustrate their signs.
PhotoDisc made those kinds of pictures, and perhaps for that reason, Getty bought them out in 1997.
Gradually, the industry went increasingly downmarket. iStockPhoto, founded in 2000, and Shutterstock, founded in 2003, focused on selling low-cost images that were priced with the internet’s scale in mind.
iStockPhoto, now known as iStock, was a pivotal site in the success of stock photography in the internet age, as it stumbled upon a micropayment model years before other companies did. And it was a total accident. In a 2007 interview with CNET, founder Bruce Livingstone noted that the firm originally went with a barter system—allowing photographers to exchange images with one another—before a huge hosting bill required the site to change to a 25-cents-per-image pay model.
“We started charging in 2001,” Livingstone recalled. “Probably toward the end of 2002 or the beginning of 2003 we actually had a bit of a budget and realized, ‘Hey, there’s a business here.’”
That business would be eventually acquired by Getty Images, but not before directly inspiring a sea of competitors, some of which didn’t charge at all. Among these was Flickr, which had embraced a Creative Commons model, allowing end users to distribute images for free in exchange for following a set of requested rules. That service has seen a variety of owners in recent years, most notably Yahoo.
(Livingstone would later leave Getty Images, but get back in the game with Stocksy, a higher-end service with bigger payouts for creators and images with a more authentic visual style.)
Over time, Getty Images consolidated power in the stock photo space, acquiring numerous competitors and licensing Corbis’ massive stock-photo library in most parts of the world, after that firm was acquired by a Chinese company in 2016. (Corbis is now known as BENlabs.) By that time, Getty’s biggest competitor emerged in the form of Shutterstock, which started as a platform closer in conceit to iStock, but has evolved to be closer to Getty’s higher-end model.
Both companies are far from alone. For example, a more recent player in the stock-photo space is Unsplash, a service that allows for free image downloads, but allows some brands to include sponsored images and offers a subscription service for higher-end photo options. These images are largely uploaded by the photographers themselves.
There’s a lot of money in stock photos—they get used everywhere, after all—but they benefit greatly from scale, size, and quality. Stock photo photographers can make as little as a few cents for a small-photo license, or $100 or more for an “extended license,” which removes usage limits and allows images to be resold in some cases.
To put it all another way, we’ve come a long way from a couple of suitcases loaded with photos.
Why the Stock Photo Model is At Risk of Disruption
Many of the major complaints lobbed at generative AI—that it’s a shortcut, that it’s used by people to displace the work of others, and that it often leads to a lower-quality end result—are often lobbed at low-end stock photos. They are basically direct competitors.
But because they are direct competitors, generative image tools are well-positioned to poke holes into the stock photo model. And stock images clearly have some problems. One of the most common ones might be the distinct “look” that stock photos often have. Stock photos, generated in photo shoots and relying on models, often don’t feel like real life, especially because the same models and settings appear repeatedly.
During the early months of the pandemic, for example, a handful of people wearing headsets and clearly looking like they were on Zoom became commonplace throughout the Web. And some stock photo models, such as Hungarian engineer András Arató, became famous memes—overexposing them in the public square while diminishing the intended value of their stock photos.
This is even the case with Creative Commons or Unsplash photos. (How many times have you seen this guy holding his head while using his laptop? His image has been downloaded nearly 300,000 times from Unsplash.)
While stock photo services like Stocksy and Unsplash have specifically taken steps to ensure their libraries have much less of an obvious “stock photo” look, stock photos are like pop music—if a pop song becomes overexposed, it starts to sound less unique over time, or can feel outdated as tastes change. Photographers can only do so much with lighting and framing to prevent a stock photo from losing its luster.
When low-end stock photos become too popular, it diminishes their value because our tastes change. While historic photos might gain value over time, many stock photos will gradually see their values shrink.
Could Stock Photos Coexist With Generative AI?
Given all that, it’s understandable why generative AI could be seen as a real alternative to stock photos in many cases. While generative AI may run afoul of shifting tastes eventually, AI will be better positioned to adapt to newer generations and styles than a stock photo that will remain wedded to the creative decisions of the time that it was taken.
(On the other hand, check out some of the AI-generated images of isolated objects in this piece—a CD-ROM, a camera, and a GPU. In each case, they look like stock photos from a distance … until you eye the details.)
Of course, a MidJourney or DALL-E image will never replace the iconic photographs in the Bettmann Archive. At best, generative AI is just interpreting a prompt based on training data; its source material is not something that came to life through the lens of a camera. But for the sea of pictures that often flood low-cost stock photo sites, generative AI creates a real risk that stock photo sites may only have two real ways to combat—either they can beat ’em, or join ’em.
Getty Images, in its lawsuit against Stable Diffusion, leaned towards the idea of beating ’em—which makes sense, because the models were most likely trained on copyrighted photos owned or licensed by Getty, which Getty says that other companies properly license.
But the join ’em approach seems to be the winning formula for Shutterstock, which actively promotes “an ethical AI-generation model that pays artists for their contributions.”
In the long run, stock photo services may be at risk, but right now, they hold lots of potential, because they often maintain archives of millions of photos that can be used as starting points for more perfectly tailored images than one might be able to find with a stock photo search.
Graphic designers know the pain of trying to source the right image to match a specific design or style. While not perfect, generative AI tools make possible something that stock photo services cannot in their original form—the ability to get a variation on an idea that may be a closer match to what the end user wants.
Stock photo photographers attempt to keep up with times and tastes, but they can’t adapt to the needs of a creator in real time, unless they’re actually working with the creator, at which point you might as well hire a photographer outright. And for that reason, it’s inevitable that generative AI will disrupt the role of the stock photo. While not a perfect product, it has numerous advantages that will be hard to ignore over time.
But there’s no reason why generative AI and stock photos can’t coexist—as long as everyone is willing to work together to find common ground.