Apple Unveils Revolutionary Text-to-Image AI for Effortless Photo Editing

Apple has released MGIE, a powerful new AI model that enables text-to-image editing capabilities for making detailed pixel-level modifications to photos through natural language instructions.

Summary

MGIE is an open-source AI model from Apple that performs image editing by understanding text prompts.
It makes use of large language models to interpret instructions and generate visual outputs.
Users can make various common Photoshop adjustments like cropping, filtering as well as advanced edits like object manipulation and style transfers simply by providing textual commands.
MGIE translates instructions into unambiguous low-level editing steps to precisely modify images.
It performs both global adjustments and localized edits to specific regions and objects in an image.
The model optimizes different visual attributes like color, texture, shape based on text prompts.
MGIE code and models are available in a GitHub repository for developers.
There is also a web demo to try out the model's capabilities.
MGIE shows the potential of using language models for intuitive image editing through natural instructions.
It represents an advance in multimodal AI and could enhance creative workflows.