Wednesday, October 9, 2024
HomeTechApple Introduces the MGIE AI Image Editing Tool, Which Utilizes Text Prompts...

Apple Introduces the MGIE AI Image Editing Tool, Which Utilizes Text Prompts to Perform Detailed Edits

Apple Introduces the MGIE AI Image Editing Tool, Which Utilizes Text Prompts to Perform Detailed Edits - The Hard News Daily

An image editing application driven by artificial intelligence (AI) dubbed MGIE was revealed by Apple researchers. It can modify photographs with basic language commands. adjustments similar to Photoshop, as well as global optimization and local adjustments, are all within MGIE’s purview. The AI tool was unveiled in the wake of Apple’s quarterly earnings call, where the tech company revealed it has been investing a “tremendous amount of time and effort” into generative AI. When compared to other artificial intelligence editing tools, the picture editing model is noticeably superior.

In order to create the tool, researchers from Apple and UC Santa Barbara worked together. As reported by VentureBeat, the research was delivered at the 2024 International Conference on Learning Representations (ICLR). The study paper is also available as a preprint on arXiv.

In the same way that Photoshop can crop, resize, rotate, and apply filters, the AI tool can accomplish the same things. On a global level, it may optimize the image by changing its contrast, sharpness, color balance, and brightness, and it can even add generative features. On top of that, it has the capability to make local modifications, where it changes or removes just one item or element within the image.

Users can easily make edits by just typing plain text commands like “make the sky brighter” or “make the house bigger.” These commands are then translated into picture commands and utilized to set the desired brightness percentage or house size metric. Users also have the option to make more detailed and intricate modifications, including “enhance the details of the leaves and the tree trunk by adjusting the dark and light areas.” A more specific request will yield a more accurate response.

There are currently AI-powered picture editors like Luminar Neo, Photoshop’s Generative Fill, Canva’s Magic Design, and FireFly (which is still in beta), but all of these programs still necessitate human intervention, either to pinpoint the exact area to edit or to apply fine-grained adjustments. However, Apple’s MGIE can do the editing task independently. A novel approach to AI frameworks enables its usage of “instruction-based image editing” or “text-guided image editing.”

The AI model employs the diffusion model, a more sophisticated architecture for realistic picture production and instruction adherence, rather than the Generative Adversarial Network (GAN) framework. The researchers then moved on to testing the model’s ability to convert text into visuals using a multimodal large language model, making sure it produced the expected results. The approach also included human assessors who ranked the adjustments and provided comments that was utilized to refine the model.

As an open-source project on GitHub, the tech giant’s MGIE AI image editing tool is now available for download. Whether Apple intends to incorporate this technology into its smartphones is currently unknown. Although rumours have it that Apple is working on new AI-powered capabilities for the iOS 18 upgrade, which is slated to arrive later this year, Apple CEO Tim Cook has stated that the company will disclose generative AI features later this year.

Author

RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Most Popular

Recent Comments