Meta has recently expanded its technology known as Expressive Media Universe (EMU) to include two new tools for image editing and video generation. This expansion comes after the initial presentation of EMU at the Connect event in September, where a new tool for creating ‘stickers’ in messaging applications was introduced. These two additional tools focus on image editing and video generation from descriptions and images.
The first new tool, EMU edit, allows for precise control of image editing through text instructions. This tool utilizes generative AI to optimize the editing process and improve precision, ensuring that pixels in the input image that are not related to the instructions remain intact. It also incorporates computer vision tasks as instructions for image generation models and has been trained using the largest dataset of its kind to date.
The second tool, EMU video, facilitates the generation of videos from a text description using diffusion models. This approach uses only two diffusion models to generate videos at 16 frames per second. The architecture of this tool responds to inputs of different types, including text-only, image-only, and text and image combined.
These new tools represent an advancement in Meta’s technology and highlight the company’s commitment to innovation in content creation and editing. According to Meta’s official blog, these tools are designed to streamline and enhance the image editing and video generation processes, providing users with new and sophisticated ways to create and edit content.