Meta has unveiled a new AI technology titled Make-a-Video which takes text prompts and turns them into videos. It’s similar to the popular DALL-E and Midjourney AIs, which utilize machine learning algorithms and a data base of online artwork to create an image based off of a text prompt.
In a Meta blog on Thursday, Mark Zuckerberg wrote that Make-a-Video is ”a new AI system that lets people turn text prompts into brief, high-quality video clips.”
It’s similar to Make-a-Scene, which was unveiled by Meta back in July. Make-a-Scene takes text prompts and outputs images.
A research paper written by a team at Meta stated that the process of teaching the AI was faster due to building it off of existing work like OpenAI’s DALL-E. Meta took image synthesis data, which are still images trained with captions, and applied video training data. This allowed for the model to learn about where a text or image prompt might exist.
Meta has not announced when Make-a-Video will be available to the public for use, but Zuckerberg did confirm that Make-a-Video will be an open-source project.
In order to combat potential dangers AI videos may present, Meta has put in a watermark in generated videos in order to ”help ensure viewers know the video was generated with AI and is not a captured video.”
Want more DZSH with you? Sign up for our newsletter today for fun games, updates, and more!