June 18, 2024
Shutterstock debuts an AI image editor for its 750-million picture library


VentureBeat presents: AI Unleashed – An exclusive executive event for enterprise data leaders. Network and learn with industry peers. Learn More


Those of us in the news media business are typically quite familiar with Shutterstock. As one of the primary large repositories that publications use to obtain stock images to illustrate articles — the others being Getty Images and Adobe Stock — it is for many publications considered critical infrastructure.

Not to mention, since its launch 20 years ago in New York City, Shutterstock has gained a userbase of multinational corporations and enterprises large and small, who rely upon stock images to illustrate marketing collateral and their web presences. The company went public on the New York Stock Exchange in 2012.

But with the advent of generative AI text-to-image models such as Adobe Firefly 2, Midjourney, OpenAI’s DALL-E 3 in ChatGPT and Bing Image Generator, Ideogram, Stability AI’s Stable Diffusion and others, the question becomes: what kind of future do stock image services have if their customers can generate custom, realistic imagery on demand using other tools?

Today, we have an answer from Shutterstock: The company is peeling back the curtain on its new AI image editing capabilities, built right into the Shutterstock website using OpenAI’s prior image generating AI model DALL-E 2.

Event

AI Unleashed

An exclusive invite-only evening of insights and networking, designed for senior enterprise executives overseeing data stacks and strategies.

 


Learn More

The new capabilities allow even free trial users of the Shutterstock service to add elements to existing Shutterstock library stock photos, extend their borders and the content along with it, change elements within them such as the colors and placements of items, remove items, switch aspect ratios, and even apply text and shapes atop them for sharing on social networks and video platforms such as YouTube.

“We’re allowing you to make stock your own and personalize it,” said Tiffany Gilron, Director of Product Marketing at Shutterstock, in a video conference interview with VentureBeat. “So our library of 750 million assets essentially becomes infinite, and whatever’s in your head you can really find it on Shutterstock.”

The new features Shutterstock is rolling out include:

  • “Magic Brush: Magically modify an image by brushing over the areas you’d like to change and simply describing what you want to add, replace or erase
  • Variations: Generate alternate options of any stock or AI-generated image
  • Expand Image: Broaden the view of any image, as easily as if zooming out through a camera lens, to see more of the scene behind the central image
  • Smart Resize: Automatically change the shape of your image to match the dimensions you need
  • Background Remover: Remove or replace the background with any scene when the subject of an image is perfect, but the background is not
  • AI Image Generator: Launched in beta earlier this year and soon to be updated with the latest version of Dall-E, this tool allows anyone to create high-quality, ethically-sourced visuals in seconds (ready for licensing and indemnifiable for commercial use) by simply describing what they are looking for.”

In the demo provided to VentureBeat, Shutterstock representatives cautioned that the features were still in beta, and it often took several seconds to generate edits — similar to the wait times found on competing AI image generators and editing platforms. However, the zoom-out feature on a stock image of a couple on a beach produced a horrifying glitched out humanoid figure in one instance, showing the limitations of Shutterstock’s current reliance on DALL-E 2.

“We’re also actively always exploring different partnerships and models,” said Kareem Isa, Principal Product Manager, during the demo with VentureBeat. “It’s such a fast-changing industry, as new models come to light, we’re consistently evaluating them. We just want what is going to consistently deliver the best output.”

Where the new Shutterstock AI image editing features live and how to use them

The majority of these features are available by clicking a black button labeled “Edit” that appears in the lower right hand corner of all available imagery on the Shutterstock library.

Clicking this button drops the user’s chosen photo into a virtual “dark room,” placing black borders around the image and bringing up a right-hand sidebar of several of these editing tools.

The user can then select which ones they want to use on the image, and for those that offer it, the features will open a text box where the user can type in the changes they’d like to make and wait while the AI feature generates several different versions to choose from.

Many of the features are designed to be “one click,” so the user simply has to select a button and the AI will spit out a few different options of what it thinks the user might want.

Some allow the user to select different colors to apply to the image from a digital palette. These either change the colors of objects within the image — such as switching an umbrella from yellow to red to match a brand’s colors, which Shutterstock demoed for VentureBeat over the videoconference — or allow the user to apply colorful borders and text blocks for title cards and brochures. The user can even select between a range of pre-set fonts for whatever text they wish to add.

The goal is to make it easier for users — from publications to brands — to not only find images that suit their needs on Shutterstock, but manipulate them directly on the platform to fit their ultimate intended uses, be they in digital or printed collateral, on social or a website or as the lead-in to a video.

“Even with 750 million assets, it’s very possible you find something that’s almost right [in the existing Shutterstock library], but not exactly what you need,” Gilron said. “Perhaps you need more white space so you can put a header and a CTA [call-to-action]. Or you need a photo for winter and the person is not wearing a winter hat and you want to add that to the image. Or the colors of the image don’t match your brand colors.”

In this way, Shutterstock is taking direct aim at other popular image editing and graphic design programs such as Adobe Creative Cloud and Canva.

Ethical AI?

But Shutterstock’s approach towards integrating AI differs from its rivals in some key areas.

For one thing — the new AI features are largely restricted to being used on existing Shutterstock imagery provided by the company’s paid creative community, rather than creating whole new images.

In fact, after a user selects and edits an image from Shutterstock’s library using the new AI tools, Shutterstock provides them with a ZIP file download of both the original source image and the one altered by AI at the user’s behest.

In this way, ” the contributor gets paid just as if you download a normal image from Shutterstock,” Gilron explained, either from an a-la-carte image purchase or through Shutterstock’s subscription plans (the company offers both options).

“We’re very strong on responsible, ethical AI,” Gilron told VentureBeat. It’s an important stance to take, especially as AI companies and those who rely on them for features face lawsuits from creators and publishers over using their work to train AI models without compensation. Even those companies that technically are allowed by their prior terms-of-service to train new AI features on existing contributor imagery and data — such as Adobe — have faced blowback and criticism from their creative communities.

What if you are a Shutterstock contributor who wants to upload an AI-generated or altered image back to the library to license out and get paid for? Sorry, you’re out of luck. The company today clarified “AI-generated or edited content will not be accepted as a submission for licensing on the platform to further ensure the protection of contributor IP and proper compensation of artists.”

Company spokespersons also told VentureBeat that Shutterstock has not and does not foresee turning on AI editing on editorial images — those taken of newsworthy figures such as politicians and celebrities, or of news events around the globe — to reduce the possibility of AI-generated or altered images being used for purposes of disinformation and discord.

Shutterstock is a member of the Content Authenticity Initiative, a trade group dedicated to promoting transparency and truth around digital content founded in 2019 by Adobe and which now consists of many news media outlets and other companies.

As Shutterstock wrote today in its news release on the new AI features, “Shutterstock intends to integrate the CAI’s underlying Coalition for Content Provenance and Authenticity (C2PA) standard into its AI capabilities and various creativity tools.” The C2PA standard, released by the separate standards working group Coalition for Content Provenance and Authenticity — also founded by Adobe and Microsoft — is a method of watermarking imagery and other content to ensure it is trustworthy and has not been tampered with.

Shutterstock’s stock price was up slightly today, less than 1% on the news.

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.



Source link