Apple unveils AI model for image editing with text commands

Apple has introduced an open-source AI model, named MLLM-Guided Image Editing (MGIE), designed to enhace image editing through text-based commands.

Developed in collaboration with researchers from the University of California, Santa Barbara, MGIE utilizes multimodal large language models (MLLMs) to interpret textual instructions for manipulating images.

The model's advanced capabilities allow users to issue simple or ambiguous text prompts, which MGIE then translates into detailed editing instructions. For example, a command to "make a pizza healthier" might prompt MGIE to add vegetable toppings to a photo of a pepperoni pizza.

In addition to major alterations, MGIE can perform tasks such as cropping, resizing, rotating, and adjusting brightness, contrast, and color balance, all through text inputs. It can also target specific areas of an image, such as modifying facial features or removing background elements.

While Apple has released MGIE on GitHub and provided a demo on Hugging Face Spaces, its future integration into Apple products remains uncertain.