Meta Unveils New Generative AI Projects for Video Creation and Image Editing
Meta, the parent company of Facebook and Instagram, has unveiled two new generative AI projects that aim to make it easier for users to create videos and customize image edits. The first project, called “Emu Video,” allows users to generate high-quality video clips based on text prompts or still image inputs. The second project, called “Emu Edit,” enables users to make custom edits within visuals by providing text instructions. Both projects are based on Meta’s “Emu” AI research project and have the potential to revolutionize creative options for brands and individuals.
The “Emu Video” project uses a factorized approach to video generation, which involves generating images conditioned on a text prompt and then generating video conditioned on both the text and the generated image. This approach allows for efficient training of video generation models. The generated videos are 512×512 pixels in size, four seconds long, and run at 16 frames per second. In human evaluations, the video generations were strongly preferred over prior work in terms of quality and faithfulness to the text prompt.
The “Emu Edit” project allows users to make specific edits within visuals by providing text instructions. The system understands the instructions and edits only the specified elements of the input image, leaving other areas untouched. This enables more powerful and precise editing capabilities, making it easier to create customized variations of visuals.
Both projects have the potential to offer new creative opportunities for brands and individuals. While Meta has not provided a specific timeline for the availability of these tools, they are expected to be launched soon. The projects demonstrate Meta’s commitment to advancing generative AI technologies and providing users with innovative and customizable content creation options.