Apple has released a new open-source AI model called MGIE, which can edit images based on natural language instructions. Developed in collaboration with researchers from the University of California, Santa Barbara, MGIE leverages multimodal large language models (MLLMs) to interpret user commands and perform pixel-level manipulations. The model can handle various editing aspects, including Photoshop-style modification, global photo optimization, and local editing. MGIE integrates MLLMs into the image editing process by deriving expressive instructions from user input and generating visual imagination for desired edits. It can perform a wide range of editing scenarios, from simple color adjustments to complex object manipulations. MGIE is available as an open-source project on GitHub, with code, data, and pre-trained models provided. Users can also try out MGIE online through a web demo hosted on Hugging Face Spaces. MGIE is a breakthrough in instruction-based image editing, demonstrating the potential of MLLMs and opening up new possibilities for cross-modal interaction. It is a practical tool for various purposes, empowering users to create, modify, and optimize images. MGIE showcases Apple’s growing prowess in AI research and development. While there is still work to be done to improve multimodal AI systems, the release of MGIE indicates the rapid progress in this field and the potential for assistive AI to become an indispensable creative sidekick.
