Introducing Apple's New AI Image Editor
There's a new Apple image editor in town, developed in collaboration with researchers at the University of California, Santa Barbara. This cutting-edge tool allows users to edit photos and images using text-based instructions. While it has not been officially released yet, you can experience a demo firsthand as reported by Extreme Tech.
The Innovation: Multimodal Large Language Model Guided Image Editing (MGIE)
Apple's latest project, known as Multimodal Large Language Model Guided Image Editing (MGIE), sets itself apart in the realm of AI image editing. Unlike traditional AI image editors currently available on the market, such as Photoshop and OpenAI's DALL-E, MGIE incorporates an additional layer of AI interpretation. This innovative approach enhances the user experience by providing more explicit and descriptive instructions. According to a paper published on arXiv, the researchers behind MGIE emphasize the significance of expressive instructions in achieving superior results.
Availability and Demonstration
For tech-savvy individuals, Apple has made an open-source version of the software accessible on GitHub. Alternatively, the researchers have set up a demonstration tool on Hugging Face. While the tool may experience some slowdown during peak usage, it presents a fun and engaging experiment for users.
Exploring the Capabilities
During a test run at the Gizmodo office, a picture was uploaded for editing using the MGIE tool. The results varied, showcasing both impressive outcomes and occasional discrepancies from the original instructions. As with any AI image generator, familiarity with the subject matter tends to yield more accurate results. The tool's interpretation of the input can sometimes lead to unexpected and creative outputs.
Conclusion
While the future of the MGIE tool's official release remains uncertain, Apple's foray into AI image editing represents a significant step towards enhancing user experiences in the digital realm.
