There is a new Apple photo editing tool on the horizon, accessible to those who are informed. The technology behemoth collaborated with experts from the University of California at Santa Barbara to create a platform that enables users to modify photos and images using text-based commands. While not officially launched yet, a demo of the tool has been unveiled by the researchers, as initially observed by Extreme Tech.
Dubbed Multimodal Large Language Model Guided Image Editing (MGIE), this project signifies a significant advancement in AI-driven photo editing. Unlike existing AI photo editing tools such as those found in Photoshop or OpenAI’s DALL-E, which facilitate image creation and editing from the ground up, MGIE aims to address the challenge of interpreting specific user inputs more effectively. According to a publication on arXiv, the researchers emphasize the importance of “expressive instructions” in instruction-based image editing, with MGIE demonstrating notable enhancements in this domain.
Apple has released an open-source iteration of the software on GitHub. Additionally, the researchers have made the application available on Hugging Face, although its performance may lag under heavy usage, relegating it to more of an intriguing experiment at this juncture.
While the MGIE application shows promise, its official release remains uncertain. Major tech corporations like Apple often back projects that may never see the light of day. Apple has yet to provide a response to inquiries regarding the tool.
The Gizmodo team took the initiative to experiment with the tool themselves. They uploaded a photograph of their colleague, Kyle Barr, sporting distinctive sunglasses acquired at this year’s Consumer Electronics Show. The AI was directed to portray the individual in a field. The outcome produced by the MGIE application envisioned the individual in a desolate setting sporting an “armored copper hat” amidst a parched and barren landscape.
This trial underscored the common limitations shared by AI image processors. The outcomes can be unpredictable and deviate from expectations. Nonetheless, the tool exhibited potential in certain scenarios, particularly when handling familiar subjects. The interpretation of Kyle’s unique sunglasses, however, posed a challenge for the AI’s accuracy.
