Getty Images and NVIDIA have partnered to launch a new tool, Getty Images’ Artificial Intelligence Generator (Generative AI by Getty Images), which for the first time allows people to create images using only Getty Images’ authorized image library.
GettyImages' GenerativeAI is trained exclusively on the vast GettyImages library, including premium content, and users receive full copyright protection. Getty promises that this means anyone who uses the tool and commercially distributes the images it creates will be protected by the law. Getty partners with NVIDIA to use Edify models on Picasso, NVIDIA’s generative AI model library.
The company says any photos created using the tool will not be included in Getty Images and iStock's content libraries. Getty will pay creators if their AI-generated images are used to train current and future versions of the model. It will share the revenue generated by the tool, "with a pro-rata share of each file and a share based on traditional licensing revenue."
"We've listened to our customers—both excitement and hesitation—about the rapid growth of generative AI and tried to be intentional about developing our own tools," Grant Farhall, chief product officer at Getty Images, said in a statement.
The Getty tool limits the types of images users can generate. It doesn’t allow me to create a photo of Joe Biden in front of the White House, or an Andy Warhol or Jeff Koons style cat. Any tips with real people's names are prohibited. When asked to create an image of the President of the United States, I saw pictures of men and women, some of whom were people of color, in front of the American flag. The model "has no idea who Andy Warhol, Joe Biden, or any other real-world person is" because it doesn't want to manipulate or recreate real-life events.
Users can access GettyImages’ GenerativeAI through the GettyImages website. The tool will be priced separately from a standard GettyImages subscription, with pricing based on tip volume, the company said. However, the company did not specify the price.
Getty Images says users will receive perpetual, worldwide and unlimited copyright to the images they create. (Still, the technical copyright status of AI-generated images remains unclear). Getty says this is similar to a customer licensing content from its repository, where the company retains ownership of the files but authorizes users to use them. Users can write their own prompts or use the prompt generator to guide them. Users can also integrate the tool into their own workflows via API. Traditionally, Getty puts a watermark on photos created with the tool to identify them as being generated by artificial intelligence.
It’s no surprise that Getty is getting into AI-powered images, as it has one of the largest image libraries. But the company has gone to war with other developers of text-to-image AI generators, suing StabilityAI for copyright infringement, alleging that its image generator, StableDiffusion, used Getty's photos without permission.
By building its own generative AI image platform, Getty could overpower other companies looking to use its library of images to train models. Getty is far from alone in using its licensed data to build an AI imaging platform. Adobe has released the Firefly model in its creative suite and creative cloud service, which is trained on its stable of licensed images.
Using copyrighted material to train large language models and text-to-image systems has been a significant concern for many in the creative community. Three artists have sued StabilityAI, Midjourney and the art website DeviantArt, accusing them of using their artwork to train models without permission.
Getty says customers will eventually be able to add their own data to train the model and generate images with their own branding. This feature and other services will be launched later this year.