Meta’s upcoming AI-based tools will let you edit any picture in some very interesting ways.
The company’s CEO, Mark Zuckerberg, teased some of the upcoming capabilities of Emu (Expressive Media Universe), Meta’s image generation foundation model revealed in September.
The first tool announced is Emu Edit, which can “precisely alter images based on text inputs.” For example, you could add a dog to an existing image, remove a person from an image, turn a dog into a panda, or change the image’s background.
Emu Video can generate “high quality” videos based on text prompts. Zuckerberg’s example show how an image of a panda can be turned into a video; other aspects of the image, including its style and setting, can also be altered.
Zuckerberg claims the company is “working on integrating these creative editing capabilities into video and photos for Facebook and Instagram.”
The demo is impressive, though it’s not the first time we’ve seen AI tools with similar capabilities; for example, Google’s Magic Editor, announced in May, can alter different aspects of a photo as well.
Google Search’s AI image generator lets you dream up products — and shop the real versions
In a blog post, Meta explained these upcoming tools in a bit more detail. The company claims that one of the advantages of Emu Edit is that it can precisely follow instructions, meaning you can alter one particular part of the image while the rest of the image stays the same. As for Emu Video, the company claims that the approach it uses generates videos of higher quality, and more faitfhul to the text prompt, than other AI tools.
The company also pointed out that Emu Video and Emu Edit are “certainly no replacement for professional artists and animators,” though it doesn’t go into details on how it plans to protect these groups from AI encroaching on their turf.