Apple develops AI image editing tool with text prompts
Apple has teamed up with the University of California, Santa Barbara, to create an AI-powered image editing tool called MGIE (MLLM-Guided Image Editing). This cool tool lets users edit photos by simply describing the changes they want in plain language. Say goodbye to complicated photo editing software, because MGIE can do things like crop, flip, resize, and add filters to images just by following text prompts.
How MGIE works and its applications
What makes MGIE unique is how it combines two functions of multimodal language models. First, it understands user prompts, and then it "imagines" the edit. For instance, if you ask for a bluer sky, the model will brighten the sky in the image. MGIE handles both simple and complex edits, like altering specific objects in a photo to change their shape/brightness. Researchers shared, "Extensive studies from various editing aspects demonstrate that our MGIE effectively improves performance while maintaining competitive efficiency."
Availability and future potential
Apple has made MGIE available for download on GitHub and also released its web demo on Hugging Face Spaces. While the company hasn't shared its plans for MGIE beyond research, CEO Tim Cook has mentioned wanting to add more AI facilities to Apple devices this year. MGIE could play a role in future vision-and-language research, helping Apple compete with big names like Microsoft, Meta, and Google in the generative AI arena.
OpenAI and Adobe also have AI image tools
Apple is not the only one with an AI-based image editing tool. Image generation portals like OpenAI's DALL-E 3, can perform quick and simple photo edits on the images they create, via text inputs. Adobe, the creator of Photoshop, also comes with its own AI editing tool. Its Firefly AI model offers generative fill. This can add generated backgrounds to images.
