The debut of a generative AI art tool that Getty Images, one of the biggest providers of stock photographs, editorial images, films, and music, calls “commercially safer” than other competing products on the market was revealed today.
The technology, dubbed Generative AI by Getty Images, was developed using a piece of the enormous (477 million assets) stock content library owned by Getty. It is powered by an AI model supplied by Nvidia, with whom Getty has tight technical cooperation. The Getty program generates photos from word prompts or descriptions of the images, similar to popular text-to-image platforms like OpenAI’s DALL-E 3 and Midjourney.
Customers who use the tool to create and download images will be granted the usual royalty-free license from Getty, which Getty claims includes indemnity (i.e., defense against copyright claims) and the right to “perpetual, worldwide, nonexclusive” use across all media.
Even though the Getty content library contains images of famous people, the company claims that it has implemented measures to stop its generative tool from spreading false information or imitating actual artists’ work. For instance, the device would not permit a user to draw a picture of Joe Biden in front of the White House or a cat in the Andy Warhol fashion, according to The Verge, which got early access to the tool. Additionally, each image the program produces has a watermark indicating that it was generated using AI.
Although Getty reserves the right to retrain its model using those photos, it states that content created by its tool won’t be submitted to its content library for others to license. Getty also says that contributors whose works are utilized to train the underlying model will receive payment. According to its statement, Getty will also split profits through the tool, providing a proportional share based on the number of files and a typical license income share.
Customers will soon be able to modify the tool with proprietary data to create photos aligned with a specific brand style or design language, and it can already be activated on the Getty website or linked to apps and websites using an API. According to Getty, pricing will depend on prompt volume and be distinct from a regular Getty Images subscription.
Grant Farhall, chief product officer at Getty, commented in a scripted statement: “We’ve created a service that allows businesses and marketers to safely embrace AI and stretch their creative capabilities while rewarding creators for inclusion of their imagery in the underlying training sets.
Getty was a vociferous opponent of generative AI tools like Stable Diffusion, which was trained on a portion of its image content database before the company released its technology. Stability AI, the AI startup that helped develop Stable Diffusion, was sued by Getty earlier this year for allegedly stealing and processing millions of photographs and related metadata that belonged to Getty without the contributors’ knowledge or payment.
Peters has, in the past, drawn parallels between the current legal climate in the generative AI field and the early years of the digital music industry, when businesses like Napster provided well-liked but unlawful services before new agreements were reached with license holders like record labels. “We think similarly that these generative models need to address the intellectual property rights of others; that’s the crux of it,” he said in an interview with The Verge in January. And to obtain clarification, we are taking [legal] action.
Some businesses creating generative AI tools, like Stability AI, claim that their methods of content exploitation are covered by the fair use theory, at least in the United States. But this issue is expected to be resolved only in a while.
It’s important to note that Getty is not the only business researching “safer,” more moral methods for generative AI (in the commercial sense).
A generative AI art tool is provided by AI startup Bria. It is trained on content that Bria licenses from partners, such as individual photographers and artists, media corporations, and stock picture repositories, in exchange for a cut of the business’s revenue. Ascendant Art, a recently established maker of avatars, on the other hand, has pledged to pay royalties to the artists who willingly submit their works to train its models.
Beyond startups, it exists. Rival to Getty Images, Shutterstock pays artists whose work is used to build artificial intelligence art models. On the other hand, Adobe claims that it is creating a remuneration structure for contributors to Adobe Stock, its repository of stock images, that will enable individuals to “monetize their talents” and gain from whatever income its Firefly generative AI engine generates.
