Apple Inc. has launched a cutting-edge multimodal AI tool named MGIE, designed for image manipulation. This innovative model allows users to simply articulate their desired outcomes, and the AI will execute the necessary modifications accordingly.

With MGIE, users have the flexibility to add or remove elements, adjust objects, enhance contrast, or create the illusion of increased lighting in images. The AI also incorporates a playful element; for instance, requesting a “healthier” pizza prompts the AI to adorn it with vegetable garnishes. Interested individuals can access MGIE by downloading it from GitHub and experimenting with it on the Hugging Face Spaces platform.

Prior to MGIE, Apple released another multimodal language model known as Ferret, which processes both textual and visual content. Unlike MGIE’s launch, Ferret’s debut did not involve any formal announcements from Apple. Ferret possesses the capability to analyze images, identify and categorize objects within them, and then highlight these objects with bounding boxes.