Apple Unveils Revolutionary AI Photo Editor Controlled by Text
Apple researchers have developed an AI image editing tool called MGIE that allows users to make edits to photos by typing text prompts describing the edits they want to make.
Summary
- MGIE stands for "MLLM-Guided Image Editing" and uses a combination of multimodal language models to interpret text prompts and imagine/generate the described edits.
- It can perform common edits like cropping, resizing, flipping images, adding filters, adjusting brightness/contrast.
- MGIE can also modify specific objects in images based on text prompts e.g. adding vegetables to a pizza image when prompted "make it more healthy".
- Apple has made MGIE openly available on GitHub to download and also provided a web demo on Hugging Face.
- It shows potential for text-based image editing without needing expertise with photo editing software.
- Apple is exploring AI generative models though isn't a major player yet compared to Microsoft, Meta, Google.
- But Apple CEO Tim Cook has expressed plans to incorporate more AI into Apple products this year.
- Comes after Apple released open-source MLX machine learning framework in Dec 2022 to train models on Apple chips.