If a few months ago Sora, OpenAI’s video creation model, made headlines for weeks, now it’s the turn of Movien Gen, the answer developed by Meta and announced today by Mark Zuckerberg’s company.
Movie Gen uses text input to Automatically generate new videos, as well as edit still images or existing footage. Audio added to videos is also generated with AI, matching images to ambient noise, sound effects and background music. Videos can be generated in different aspect ratios.
In addition to generating new videos, Meta says Movie Gen also You can create custom videos from images or take an existing video and change different elements of it. An example shared by the company shows a still photo of a woman; The added video shows her sitting in a pumpkin patch sipping a drink.
Movie Gen can also be used to Edit existing footage and change styling and transitions or add things that didn’t exist before. In an example shared by Meta, a relatively innocuous video of what appears to be a runner is edited using AI in different ways: In one frame, he is holding pom-poms. In another, the background has been edited to represent a desert. In a third, the runner is wearing a dinosaur costume. Changes can be made using text prompts.
Meta chief product officer Chris Cox said the company “not yet ready to launch Movie Gen as a product in the near future”, as it is still expensive and the generation time is too long.
Little else is known about this AI, how you trained, how long it takes to generate the videos, where you get the audios from. For now, Meta has only said that it trained Movie Gen with “a combination of authoritative and publicly available data sets,” but did not specify which ones.