Apple Intelligence Brings Generative AI to Photo Editing

Three smartphone screens show an image editing app with a photo of a family standing on a rocky beach with cliffs in the background. The app's interface displays editing tools and highlights options to remove objects and people from the photo.

At Monday’s WWDC keynote, Apple unveiled new artificial intelligence (or as the company is referring to it, Apple Intelligence) updates, making good on the company’s earlier hints.

Unsurprisingly, photos and images were included in the tech company’s round-up of AI news.

Apple doesn’t seem inclined to go full Dall-E on the image generation front. Its Image Playground tool forces users to choose between three styles: animation, illustration, and sketch, so there are no deepfake possibilities with photorealistic AI generation. Alongside a dedicated app, the feature will be deeply integrated into other apps, including Messages, Notes, Pages, Keynote, and Freeform. Apple will also extend access to the tool to third-party apps.

As presented in the keynote, Image Playground feeds off the context it is surrounded by. In Messages, Apple highlighted that this could mean an animation-style image of a friend surrounded by balloons and a cake when you wish them a happy birthday. Image Playground can also take what it knows about how the people in your life look and depict artsy versions of them in various settings, with different accessories, or amid new locations.

Additionally, users can now customize entirely new emoji creations with Apple’s new “Genmoji” feature.

This tool is further enhanced with the Image Wand, which allows users to circle a subject and artificially generate an image based on the surrounding context. As Apple showed in an example, this can even turn a simple sketch into a more elaborate visualization within a Notes page.

The Photos app is also due to receive significant Apple Intelligence upgrades in an attempt to keep up with the AI other smartphones are now packing. With Clean Up, distracting elements of an image’s background will automatically be identified and removed, for example. Presumably, this requires some level of pixel generation in photos. After all, a removed object must be replaced by something. However, Apple does not appear to be enabling text-to-image prompts when editing specific images.

The AI model can also use more conversational search prompts to find relevant images. It can even work to find the right spot in a longer video. This pulls away from the current search function in Photos, which relies on basic keywords like “cat” or “wedding dress” or people already identified. In Apple’s example, the company showed the result of a search prompt asking for an image of a specific person with stickers on her face.

Similarly, the Memories feature will allow users to give general prompts, and Apple Intelligence will pull several relevant photos in a video format with suggested audio from Apple Music.

These updates are pretty similar to what competitors are offering and as clear a sign as any that Apple is not willing to be left behind in the AI arms race.


Image credits: Apple

Discussion