Apple Releases Open-Source AI Image Editing Model
Apple has entered the realm of AI image editing with the release of an open-source multimodal AI model.
What is Apple Planning?
Earlier this week, a collaboration between Apple and the University of California, Santa Barbara led to the unveiling of MLLM-Guided Image Editing (MGIE). This AI model enables image editing akin to Photoshop through simple text commands.
Apple’s Approach to AI Development
In the arena of AI technology, Apple has maintained its characteristic discretion regarding its plans. Despite the fervor surrounding last year’s ChatGPT trend, Apple refrained from making significant AI announcements. However, reports suggest the company has been working on an in-house chatbot akin to ChatGPT, known as “Apple GPT,” with CEO Tim Cook hinting at forthcoming major AI revelations later this year.
Innovative Image Editing with MGIE
While various AI image editing tools exist, they often struggle with concise human instructions, resulting in subpar outcomes. MGIE offers a novel solution by leveraging multimodal large language models (MLLMs) to comprehend text prompts and image data effectively. This approach eliminates the need for exhaustive descriptions.
Examples of MGIE’s Capabilities
Demonstrations from the research showcase MGIE’s proficiency. For instance, given an image of a pepperoni pizza and the directive “make this more healthy,” MGIE deduces that adding vegetables would fulfill the request, resulting in a pizza adorned with greenery.
Similarly, when tasked with adding lightning and ensuring its reflection on water in an image of a forested shoreline, MGIE excels compared to other models, successfully incorporating the lightning reflection.
Accessibility of MGIE
Interested parties can access MGIE as an open-source model on GitHub or as a demo version hosted on Hugging Face.
Grow your business with AI. Be an AI expert at your company in 5 mins per week! Free Newsletter – https://signup.bunnypixel.com