• Accessible on the iStock website, users can instruct the new tool to generate a wide range of content.
  • The tool, trained on Getty’s extensive library of licensed images, uses generative artificial intelligence to turn user text descriptions into vivid visuals.

Getty Images Holdings Inc., a leading American visual content marketplace for videos and images, recently unveiled its latest innovation, Generative AI by iStock, at the CES 2024 consumer electronics show. This tool empowers creators to generate commercially safe content using licensed images.

The newly introduced tool employs generative artificial intelligence capable of transforming user-provided text descriptions into vibrant visuals. Trained on Getty’s extensive library of commercially licensed images, the AI tool ensures that legal protection and usage rights for content creators accompany any content it generates.

Nvidia Picasso, a cutting-edge cloud service from Nvidia Corp., powers the AI underneath and can convert text to images in a manner akin to well-known AI models like OpenAI’s DALL-E. It can produce a wide range of creative visuals in response to user commands by being trained in collecting images, graphics, and videos from Getty Images.

With imaginative prompts like “An astronaut bunny painting the moon with intergalactic colors” or more realistic ones like “A businessman in a grey suit sipping coffee in a bustling Tokyo café, surrounded by modern architecture.” users can ask the new tool, which is available on the iStock website, to create almost anything. The tool produces photos with a resolution of up to 4K and photo quality.

Developers can soon access advanced image editing capabilities through application programming interfaces (APIs), enabling seamless integration with their creative apps. These editing features encompass both inpainting and outpainting.

The inpainting feature allows users to mask a specific part of an image and use a text prompt to describe the content they want in that area. For instance, a user might select a circle on a branch in an image of a rainforest and instruct the AI to ‘Add a white bird.’ The model will then generate a small bird with white feathers in the designated area.

With outpainting, users can enlarge an image’s edges and then have the AI create fresh content based on the original image for the expanded area. With the help of this feature, users can easily increase the size of their images both horizontally and vertically without compromising their original appearance. For instance, consider an image of a car cropped on one side—users can use the AI model to extend it, and the missing sections will be automatically completed.

According to Getty Images, users will be able to swap out individual objects and people inside the masked area with a “replace” feature that functions similarly to inpainting.

Shutterstock Inc., a competing image library, unveiled its generative AI image creation tool in collaboration with OpenAI, employing the DALL-E 2 AI model in October. The image library provided images from its collection to train DALL-E’s machine learning models and seamlessly integrates the generative tools into its platform.