Robert Triggs / Android Authority
TL;DR
- Apple has co-created an AI mannequin that may carry out superior edits on pictures primarily based on textual content prompts.
- MGIE can fully alter a picture by performing edits like changing backgrounds, manipulating topics, eradicating objects, and far more.
- The AI mannequin was introduced in a analysis paper and isn’t one thing we count on to see on an iPhone anytime quickly.
Apple and researchers from the College of California, Santa Barbara, have co-created an AI instrument that’s able to performing picture edits primarily based on textual content prompts (by way of Enterprise Beat).
Referred to as “MGIE,” the AI was introduced in a paper on the Worldwide Convention on Studying Representations 2024. It’s a multimodal massive language mannequin, like Google Gemini, that may edit pictures very similar to you’d do on Photoshop. Solely right here, you possibly can specific your ideas in textual content and the AI will do all of the modifying give you the results you want.
Say you’ve a picture of a Pizza. You possibly can inform MGIE to “make it extra wholesome,” and it’ll add more healthy toppings to the pie within the picture. Apple’s co-authored paper additionally presents different edit use instances the place you possibly can take away objects from pictures, change colours, and improve lighting and different particulars of a picture. It may well even flip a forest path right into a seaside, change the background of images, create inventive sketches, and far more. Consider Google’s Magic Editor on steroids. You possibly can view examples of MGIE’s modifying capabilities right here.
“MGIE consists of an MLLM (Multimodal Giant Language Mannequin) and a diffusion mannequin. The MLLM learns to derive concise, expressive directions and provides express visual-related steerage. The diffusion mannequin is collectively up to date and performs picture modifying,” the paper explains.
There’s no telling how Apple plans to make use of these learnings on precise consumer-facing picture modifying instruments. We do know that the corporate is engaged on generative AI options for its platforms. It’s potential we would see AI-based modifying instruments on the brand new iPhone 16 sequence. Though we presume MGIE’s in depth modifying capabilities may want a wholesome quantity of processing, so Apple may introduce a toned-down model of the AI if and when it’s utilized on iPhones.
In the event you’re focused on attempting out MGIE, you possibly can try a demo hosted right here.