Thanks to the collaboration with theUniversity of California, Apple was able to release a new and interesting AI model capable of carrying out “textual photo editing“, i.e. modifications to images through simple textual input.
The model in question is known as MGIE (that is to say MLLM-Guided Image Editing) and can interact at different levels with any image. For example, through promptit is possible to intervene on specific objects present in a photo, changing the shape of the same or changing its brightness.
What makes MGIE unique is the fact that it combines the use of two multimodal linguistic models different. One of them is useful for interpreting the user’s orders and the second for “imagining” the result of the modification he is about to perform.
Regardless of these technicalities, the result of this combination is clear: even without using specific photo editing apps, which are not accessible to everyone, anyone will be able to make changes to photos and images.
MGIE could be the appetizer of what Apple has in store for its customers
As reported by VentureBeatApple has made MGIE available via GitHub for download, also releasing a web demo on Hugging Face. Despite this, at least at the moment, the Cupertino giant has not revealed any particular future projects or evolutions of MGIE.
Although this project is interesting, it must be taken into account that it is not an absolute novelty. Already other platforms, such as FROM-E 3 Of OpenAI, perform photo editing operations through text input. In this sense, it should be remembered that also Adobe applied theArtificial intelligence to image editing, with Firefly.
While not revolutionary, this new tool once again demonstrates how Apple has finally considered its entry onto the scene in the context of AI. After the confirmations of Tim Cookin fact, MGIE appears with a further step in this direction by the company.