Good morning dear readers of Tecnogalaxy, today we will talk about the artificial intelligence of Meta that allows you to transform sentences into video.

Meta presented an artificial intelligence system that generates short videos based on text messages.


Make-A-Video allows you to type a series of words, such as “A dog wearing a superhero outfit with a red cloak flying in the sky”, and then generates a five-second clip that, although fairly accurate, has the aesthetics of an old video with low quality.

Although the effect is rather crude, the system offers a preview of what will come next for generative artificial intelligence, and it is the next obvious step of text-to-image AI systems that have aroused great enthusiasm this year.

The announcement of Meta by Make-A-Video, which has not yet been made available to the public, will likely push other artificial intelligence labs to release their own versions. It also raises some big ethical questions.

Only in the last month, the OpenAI artificial intelligence laboratory has made available to everyone its latest AI system from text to image DALL-E and the launch of Stability artificial intelligence. But AI converting text to video presents some even bigger challenges. First, these models require a great deal of computing power. They represent an even greater computational increase than the large text-to-image AI models, which use millions of images for training, because putting together one short video requires hundreds of images. This means that only large technology companies can afford to build these systems for the foreseeable future. They are also more difficult to train, because there are no large-scale data sets of high-quality videos paired with text.


To overcome this problem, Meta combined data from three video data sets and open source images to train their model. Standard text image datasets of tagged still images have helped AI learn what objects are called and what they look like. And a video database helped him figure out how those objects should move around the world. The combination of the two approaches helped Make-A-Video, described in an un-peer-reviewed document released today, to generate large-scale text video.

Tanmay Gupta, computer vision researcher at the Allen Institute for Artificial Intelligence, says Meta’s results are promising. Shared videos show that the model can capture 3D shapes while the camera rotates. The model also has some notions of depth and understanding of lighting. Gupta says that some details and movements are done decently and convincingly.

However, “there is plenty of room to improve the research community, especially if these systems are to be used for video editing and professional content creation,” he adds. In particular, it is still difficult to model complex interactions between objects.

For its part, Meta promises that technology could “open up new opportunities for creators and artists”. But as technology develops, it is feared that it can be exploited as a powerful tool to create and spread misinformation and deepfake. It might make it even harder to distinguish between real and fake online content.

The Meta model increases the stakes for generative AI both technically and creatively, but also “in terms of the unique damage that could be caused by video generated versus still images,” says Henry Ajder, synthetic media expert.

“At least today, creating actually inaccurate content that people might believe in requires some effort,” says Gupta. “In the future, it may be possible to create misleading content with just a few buttons.”

A spokesperson for Meta said it is not yet making the model publicly available and that “as part of this research, we will continue to explore ways to further refine and mitigate potential risks”.

Read also:

Was this article helpful to you? Help this site to keep the various expenses with a donation to your liking by clicking on this link. Thank you!

Follow us also on Telegram by clicking on this link to stay updated on the latest articles and news about the site.

If you want to ask questions or talk about technology you can join our Telegram group by clicking on this link.

© - It is forbidden to reproduce the content of this article.