Meta announces AI tool that turns text into moving images
Meta has announced Make-A-Video, an AI tool that converts text into moving images. According to Meta, the AI model learns what the world looks like by analyzing image captions and watching video images that don’t contain text.
On the website of the Make-A-Video shows how the AI tool can generate videos when text is entered. The tool can also work with existing images to create a moving video and, according to Meta, also succeeds in making variations of videos uploaded by users.
Meta has developed filters to prevent the tool from generating malicious videos. The company will also add a watermark to each video to inform users that it is a video created by artificial intelligence. The researchers want to make the AI tool available to the general public in the future, but for now the project is still in a test phase. Researchers who want to participate in the project can apply via Meta Sign Up.
The Meta researchers shared their research through a paper. The project is according to the introduction open source. The paper explains how they plan to further train the AI model via the Make-A-Video tool. For example, they want to teach the underlying AI model what the world looks like on the basis of moving images that were not supplied by the researchers. The tool also allows them to save a lot of time and continue to see the range of visual interpretations of the AI model in action.
An example of Meta’s Make-A-Video tool. Click for a gif. Source: meta