Apple to upgrade on-device image models in iOS 27, add third-party AI image generation support

Apple plans to significantly improve the visual quality of its on-device image generation models for Genmoji and Image Playground in iOS 27, according to Bloomberg's Mark Gurman in his Power On newsletter.

Current limitations

Image Playground and Genmoji launched in iOS 18.2 with notably poor image quality compared to competing tools. While Genmoji produced acceptable results, Image Playground's output was "incredibly poor," according to 9to5Mac's analysis. Both features run on-device models, which partially explains the quality constraints.

What's changing

According to Gurman, Apple's own models will receive "a big boost" in visual quality this year. The report does not specify whether the improved models will continue to run on-device or will require cloud processing.

Apple will also expand third-party model support in a redesigned Image Playground app. Currently, the feature only supports ChatGPT image generation. The company will likely add support for Google's image generation models and other competitors, though specific partnerships have not been disclosed.

Additional features

Apple is also working on suggested Genmoji functionality that will draw from users' photo libraries and other personal data, though details on this feature remain limited.

What this means

Apple's on-device image generation has been a competitive weakness since launch. The planned improvements suggest Apple recognizes the quality gap and is willing to either improve its on-device models or potentially shift to cloud-based processing for better results. Adding third-party model support also represents a strategic shift, allowing users to choose between Apple's on-device privacy approach and potentially higher-quality cloud-based alternatives. The changes could arrive with iOS 27's release later this year, though Apple has not confirmed a timeline.