Getty is adding an AI image generator to iStock

Getty Images, the company known for selling stock imagery for digital and print content, today pushed its AI efforts ahead by launching a dedicated image generation service for its subsidiary iStock.

Dubbed Generative AI by iStock, the offering gives designers and business users access to a text-to-image generator that produces ready-to-use visuals from simple text prompts.

However, the interesting bit here is that the images generated by this tool come with legal protections and usage rights – unlike many other image generators out there.

This means a business could use the image generated by the tool for commercial purposes, like for their marketing content, with protection against the risk of it being stolen for use elsewhere.

The announcement marks Getty’s second notable move in the elusive generative AI space, following the launch of an image generator for its main platform in September. It also comes in the wake of rival stock image service Shutterstock’s launch of an AI image generator feature that can adapt images from its library of assets back in autumn 2023.

What to expect from Generative AI by iStock?

Showcased at CES 2024 in Las Vegas, Generative AI by iStock is pretty similar to what Getty debuted for its own site.

Users will get a text-to-image tool, where they can describe what they want and the underlying model will generate photo-quality images at up to 4K resolution for that.

Getty, having a massive catalog of licensed images under its umbrella, touts that this tool has also been trained on its (and iStock’s) proprietary creative library, covering exclusive photos, illustrations and videos.

This makes it an ideal offering for licensed AI content, especially at a time when artists have been suing AI tools over using their artworks for training models without permission. Getty Itself has sued Stability AI over similar concerns.

It remains to be seen whether training on stock content limits the quality of output in any way or not.

Either way, by training on a licensed library, Getty is assuring iStock customers that they will get commercially safe content in the output. Each generation will produce four distinct images and each image will come with a legal coverage of up to $10,000.

This indemnification will ensure that users are protected against the risk of their content being stolen. The company also notes that iStock will not add the generated content to its creative library for others to license.

Notably, the tool also comes with a prompt builder that gives users separate fields to define various aspects of their input, like what is the subject, what is it doing and what are their surroundings. It also includes added options to define the color and mood of the image to be generated as well as aspect ratio.

Nvidia Picasso model under the hood

The entire generation effort here is driven by Nvidia’s Picasso AI model library. Essentially, Getty combined Picasso’s Edify foundation model for visual design with its licensed content to create a bespoke model that powers the Generative AI by iStock service.

The offering is available starting today with a dedicated pricing structure for 100 generations. In addition to this, Getty is launching advanced editing capabilities for its users, including inpainting and outpainting features.

Inpainting allows users to mask a specific region of an image and replace it with an object specified via text input, while outpainting allows users to expand the image to fit various aspect ratios, filling in content based on the original image. Both features are set to launch on iStock and Getty Images. Meanwhile, developers can get APIs to integrate the capabilities into their creative applications right away.

In the much-hyped AI-driven content generation space, Getty’s AI tools go against players such as Midjourney, DeviantArt, Runway, and Stability AI. Adobe has also released its Firefly model, trained on its proprietary catalog of licensed images, across several Creative Cloud products.