Revolutionizing Visual Artistry: Getty Images’ Cutting-Edge Generative AI


Getty Images is collaborating with Nvidia to introduce Generative AI by Getty Images, an innovative tool enabling individuals to generate images from Getty’s extensive repository of licensed photographs.

Generative AI by Getty Images, though bearing an unwieldy moniker, derives its training exclusively from the expansive Getty Images library, encompassing premium content. This confers full copyright indemnification to users. In practical terms, anyone utilizing this tool to produce and commercially publish the resulting image is assured legal protection, a pledge extended by Getty. The partnership between Getty and Nvidia harnessed Nvidia’s Edify model, accessible through Nvidia’s repository of generative AI models, Picasso.

Craig Peters, the CEO of Getty, is scheduled to participate in this year’s Code conference, slated for September 26th and 27th. For those interested in attending Code in person, applications are open, or you can acquire virtual tickets here.

I had the privilege of a hands-on encounter with Generative AI by Getty Images, affording me an opportunity to experiment with the tool. My primary curiosity lay in assessing its ability to produce photographs that closely resembled authentic Getty-watermarked images. To my pleasant surprise, the resultant photographs surpassed my expectations. Stock photos inherently exhibit an artificial and somewhat soulless quality, and it did not astonish me that a few initial images generated by the tool evoked a sense of emotional vacancy. This sentiment is not exclusive to Getty’s generative AI tool; I encountered a similar reaction when assessing the output of OpenAI’s forthcoming DALL-E 3.

Getty’s tool demonstrated notable proficiency in rendering lifelike human figures. My prompt entailed the creation of a photograph featuring a ballerina poised in an arabesque stance, one leg gracefully extended behind, against the backdrop of a slightly blurred stage. The images produced possessed a more authentically human quality than those generated using Stable Diffusion under the same directive. In fact, the Getty-generated image managed to deceive my acquaintances when shared with them. It is evident that Getty’s model draws inspiration not solely from illustrated art but from actual photographs. In contrast, the tool’s illustration mode yielded solely two-dimensional, clip-art style interpretations of the same prompt.

The company has stipulated that any images generated through the tool will not be incorporated into the Getty Images or iStock content repositories. Furthermore, Getty commits to compensating creators when their AI-generated imagery is utilized to train present and future iterations of the model. Revenue generated through the tool will be equitably distributed, encompassing both a pro rata share per file and a portion derived from traditional licensing revenue.

Grant Farhall, Chief Product Officer at Getty Images, remarked, “We’ve attentively heeded our customers’ feedback regarding the rapid evolution of generative AI – responses have ranged from enthusiasm to caution. Consequently, we’ve taken a deliberate approach to the development of our proprietary tool.”

The capabilities of the Getty tool are subject to certain limitations in terms of the types of images users can generate. It imposed restrictions that prevented me from generating a photograph featuring Joe Biden in front of the White House or a cat fashioned in the stylistic vein of Andy Warhol or Jeff Koons. Any prompt containing the name of a real-life individual was categorically prohibited. A request for an image depicting the President of the United States yielded a selection of images featuring both male and female figures, including individuals of diverse ethnic backgrounds, set against the backdrop of the United States flag. The company clarified that the model “lacks awareness of real-world figures such as Andy Warhol, Joe Biden, or any other individual” to preclude manipulation or replication of real-life events.

Users can access Generative AI by Getty Images through the Getty Images website. The company has confirmed that the tool will be offered as a separate service, distinct from a standard Getty Images subscription, with pricing determined by prompt volume. However, specific pricing details have not been disclosed.

Getty has assured users of perpetual, worldwide, and unrestricted rights to the images they create. Nevertheless, the precise copyright status of AI-generated images remains somewhat ambiguous. Getty has likened this arrangement to the licensing of content from its library, wherein the company retains ownership of the file while granting permission for its use. Users may either craft their own prompts or avail of the prompt builder to guide them. Additionally, users have the option to integrate the tool into their existing workflows through an API. Consistent with tradition, Getty appends watermarks to pictures generated by the tool, thereby identifying them as AI-generated.

It comes as no surprise that Getty has entered the arena of AI-generated imagery, boasting one of the most extensive repositories of images. Notably, the company has engaged in legal battles with other developers of text-to-image generative AI, initiating lawsuits against Stability AI for alleged copyright infringement, citing the unauthorized use of Getty photos in the Stable Diffusion image generator.