Getty Images and Nvidia introduce generative AI for stock photos
Getty Images and Nvidia are deepening their partnership in artificial intelligence with the launch of Generative AI by iStock, a text-to-picture platform specifically designed for stock photography.
Generative AI by iStock is based on Getty Images’ first attempt to create artificially intelligent images – Generative AI by Getty Images. The difference is that the image platform from iStock, a stock photo service owned by Getty, helps private users or users with one workplace (small businesses), unlike Getty Images, which is more of a multi-user enterprise solution.
Trained using Nvidia’s Picasso model, iStock’s Generative AI was only trained on the Getty creative library and iStock’s stock photo library. It was not trained on Getty’s library of editorial images to prevent generating trademarks or famous personalities.
“It allows users to be more efficient in their workflow and get the more accurate photos they need, even the things they can’t do with a camera,” says Grant Farhall, chief product officer at Getty. He gave the example of a person looking for photos to illustrate climate change: they can ask iStock’s Generative AI to create a photo of penguins walking down the street; instead of hiring a photographer and looking for a flock of penguins, artificial intelligence can do it for them.
The price is $14.99 for 100 promts, each of which generates four images.
Another big difference between Getty Images’ AI platform and the new iStock service is legal liability. Unlike Generative AI from Getty Images, users will not have unlimited refunds. The iStock platform will have a limit of $10,000 per asset, the same license it offers for its current library. As with the first generative AI platform from Getty, authors whose content was used to train the model can participate in a revenue sharing program.
The iStock platform will also get Inpainting and Outpainting features “soon”, Alexander Lazarou, Getty representative, told The Verge. Inpainting allows users to mask an area of an image and then fill it with a person or object from a text prompt. Outpainting expands the photo for different aspect ratios and fills in these new areas.
