Why Getty Images and Picsart are partnering to train a new AI image model

As AI-generated content platforms continue rolling out new AI models and features, Getty Images has made another data licensing partnership with another player in AI image generation.

The stock image giant has signed a new deal with Picsart, a content creation platform, which will allow Picsart to develop a new AI image generation model trained on Getty Images’ licensed creative content. Rather than using the images to train an existing AI model, Picsart’s PAIR AI lab plans to develop a new model from the ground up and make it available for use by both companies later this year. 

Announced Thursday, the deal aims to provide “commercially safe AI-generated imagery” for creators, marketers and small businesses. Along with providing customers with new ways to create and edit images on both companies’ AI platforms, it will offer customers commercial rights and indemnity for the images they create while developing new ways to compensate the creators of the images used to train the AI model. Terms of weren’t disclosed, but Getty Images chief product officer Grant Farhall told Digiday the partnership will have recurring components and “isn’t a deal where there’s a one-time fee for using our content.”

“Picsart has a great set of tools as part of their design studio partnering with us to build AI capabilities and a model from our creative content,” Farhall said in an interview. “We’ve consistently said we are ready and keen to partner with companies who understand the value of license content and are choosing to train their models in a responsible way that respects the rights of IP holders and of artists.”

The partnership is just one of several Getty Images has announced in the past year. Along with debuting its own AI image and video creation platform last fall, Getty Images announced partnerships with the Israeli AI image platform Bria and a few months later with Brooklyn-based AI video startup Runway. Another partnership is with Nvidia, which earlier this year announced using Getty Images content to train a platform for AI-generated 3D content.

Consumers are also calling on AI content platforms to provide more transparency, according to recent report released by Getty Images. In a two-year survey of 30,000 adults in 25 countries, 90% of survey respondents said they want transparency on AI Images while 76% admitted it’s become harder to tell which images are real or not.

Using content from Getty Image will provide Picsart with a massive amount of high-quality content and allow the firm to reach a new set of potential customers. In an interview with Digiday, Picsart Chief Financial Officer Craig Foster described Getty Images as “the Ferrari of stock images.” He thinks training an AI model will help alleviate some of the ongoing concerns companies have with creating AI-generated content for commercial use.

“We are able to sell the model into our ecosystem, they’re able to sell the model into their ecosystem,” Foster said. “They have more higher-end clients than we do, which is going into ad agencies and more professional-use cases as we continue our journey up the stack a little bit … They probably have one of the best datasets in the world.”

Founded in 2011, the Softbank-backed startup now has more than 150 million monthly users and first debuted its generative AI image in November 2022. Since then, it’s added hundreds of tools powered by generative AI, which will be available to Getty Images customers. According to Foster, entrepreneurs and creators make up between 20% and 25% of its user base while about 85% of the company’s total user base is Gen Z. He also noted 125,000 new pieces of content are added to the platform every day. 

The partnership between Getty Images and Picsart could also help both companies compete in the crowded race for dominance among many generative AI content platforms. Earlier this week, Stability AI released a new AI image model called Stable Diffusion 3 Medium, which follows the release of Stable Diffusion 3 earlier this year. (Stability AI also happens to be in an ongoing legal battle with Getty Images, which sued the company last year over copyright concerns.) 

In its latest update this week, Adobe announced new features for its Adobe Express platform including expanded access for its Firefly 3 image model, ways for companies to use their own AI models, and new AI integrations with other major platforms like Microsoft Copilot and ChatGPT. Adobe recently faced scrutiny after users raised concerns about ownership of AI-generated content and the company’s use of it for training AI models. The company — which quickly said the legal terms didn’t match actual policies — then said it won’t train its AI models customers’ content. (Updated terms are expected for next week.)

Despite ongoing concerns about AI-generated content, marketers’ appetite for creating it or advertising adjacent to it continues to rise. In GroupM’s mid-year forecast released this week, machine-generated content will account for just 2% of content-driven advertising in 2024 in terms of share of ad revenue. However, it could grow to more than 10% by 2029 across all types of content including TV, film, audio, print and social. Adjacent concerns about issues like copyright and brand safety could lead marketers to call on AI content creation and distribution platforms to require more transparency.

“I think what we were especially getting at with the proportion of [machine-generated content] on these platforms is for advertisers — if they are knowingly investing in these platforms with user generated content that comes with substantial reputational risks — that these brands are signed up to,” Kate Scott-Dawkins Global President, who leads business intelligence at GroupM, said on a press call last week. 

Beyond using generative AI for content-generation, Getty Images also is investing in other features for AI editing and overhauling how customers search for images on its platform. Farhall noted it is often more efficient and effective to search through hundreds of pre-shot creative images than to generate new AI images. Rather than picking a few keywords, adding ways to search for images using natural language helps customers find what they’re looking for while using full sentences or “longer, more expressive phrases.”

“Our customers show up on our experience and on our website and they have probably got a picture on their head, but not always,” Farhall said. “…If you actually allow people to express themselves in more complete and natural ways, we probably have a better understanding of what they’re actually looking for.”

https://digiday.com/?p=547781

More in Media

AI Briefing: Senators propose new regulations for privacy, transparency and copyright protections

A new bill called the COPIED Act aims to pass new transparency standards to protect IP and guard against AI-generated misinformation.

Media Briefing: Publishers reflect on ad revenue midway through 2024 

Some publishers say ad revenue is pacing 15% up year over year while others are still managing their expectations for how 2024 will shake out.

Teads is exploring sale options as M&A in ad tech heats up

Sources state the Altice-owned stalwart of outstream video has recently held talks with private equity and strategic players.