Meta Unveils New AI Tools For Video Creation and Photo Editing
Meta announced two new AI tools for video creation and photo editing. Based on Emu, Meta’s foundational model for text-to-image generation, Emu Video and Emu Edit will roll out on Instagram and Facebook in the future.
Emu Video is a text-to-video tool that allows users to create AI videos from text or image prompts, or a combination of both. Based on the input, Emu Video first generates images and then connects them into four-second-long videos. This two-step approach is critical to train Meta’s AI video models more efficiently and supports the direct generation of high-resolution videos. Emu Video can also animate images based on text prompts.
Similarly, Emu Edit allows users to enhance photos and videos using text prompts. Described as a “free-form editing” tool, Emu Edit can assist in “local and global editing, removing and adding a background, color and geometry transformations,” and more. According to Meta, this model surpasses current AI photo editing tools in the precision of the applied edits, eliminating problems that make AI-generated images look too artificial or poorly edited.
Meta states that Emu Edit trained on a dataset of 10 million samples containing images and instructions, which is responsible for the model’s ”unprecedented edit results in terms of both instruction faithfulness and image quality.”
While the new generative AI tools from the Emu family are still in early research, Meta sees great potential in their real-life application. According to the company’s blog post, Emu Video and Emu Edit could help both users with no technical skills and content creators generate and edit beautiful visuals in seconds. Emu has a fun side, too – users in group chats can create custom “animated stickers or clever GIFs on the fly.”
The company stresses Emu Video and Emu Edit aren’t intended to replace professional artists but rather assist Instagram and Facebook users in expressing themselves in new ways.
In addition to Emu, Meta has another AI image-creation tool, the human-like I-JEPA. What sets it apart from other AI models is its ability to apply background knowledge about the world rather than focus solely on technical issues like pixels. This unique approach allows it to perform better in analyzing and completing unfinished images.
Meta has been on an AI streak. Reportedly paying millions of dollars to famous figures to use their likeness, the company also released AI personas, which are basically chatbots with a celebrity twist.
