Forget Dall-E, Meta totally generates videos with simple text

Meta presented Make-A-Video, its artificial intelligence system that allows you to create short videos from a description that you give. When DALL-E and many AIs create images, Meta positions itself on the video.

Screenshots of videos created by Meta’s AI // Source: Meta

For several months, the artificial intelligence program DALL-E 2 has met with great success on social networks. It allows to create images automatically thanks to a textual description given by the user. It’s not the only AI program that exists, but Meta has decided to go further by positioning itself on video with its Meta AI department. With Make-A-Video, the company wants to make this available to everyone in the future. Meta wants to be transparent and makes a research document public.

Create video from text

Make-A-Video does not start from scratch and relies on image-to-text technologies to generate video from text. On the project pagewe learn that “the system uses images with descriptions to know what the world looks like and how it is often described“.

Forget Dall E Meta totally generates videos with simple
“A dog wearing a superhero outfit with a red cape flying in the sky” according to Make-A-Video // Source: Meta

Then, the artificial intelligence behind this system uses videos to analyze how objects or living beings move. According to Meta, a few words or lines of text are enough to create videos. Whether surreal, realistic or stylized, the project has published several GIFs of the experiments carried out.

1664818727 959 Forget Dall E Meta totally generates videos with simple
“A teddy bear painting a portrait” created automatically // Source: Meta

Make-A-Video is also able to animate an image from one or two images. Finally, another demonstration shows that you can create variations of an existing video. For the moment, only GIFs have been shown, in a fairly low resolution and quality, but it is clear that on some examples the result is quite impressive.

Will we be able to use Meta’s Make-A-Video?

Like any artificial intelligence program that can be made available to the general public, Meta protects itself and says it is committed to “guarantee the safe use of this advanced video technology“. Make-A-Video is equipped with filters for “reduce the risk of generating harmful contentand appends a watermark to the generated videos to avoid possible hijackings.

Meta says it wants to make Make-A-Video available to the public, but hasn’t announced availability. We only know that Meta plans to “publish a demo experiment“. We can initially imagine a model like that of DALL-E 2, with a system of registration and credits. But in a longer time, Make-A-Video could be integrated directly into Meta applications (Facebook, Instagram, WhatsApp, etc.). We could ask the function to create GIFs from a description.

To follow us, we invite you to download our Android and iOS app. You can read our articles, files, and watch our latest YouTube videos.

We wish to say thanks to the author of this short article for this incredible content

Forget Dall-E, Meta totally generates videos with simple text

Check out our social media accounts along with other pages related to them