Meta has introduced its latest artificial intelligence (AI) system that can generate videos using only text prompts. The tech company announced on Thursday, September 29; the Make-A-Video AI is trained with existing images with captions and unlabeled videos to determine how the world moves.
The AI system generated videos that ranged from surreal to creepy to impressive. Meta did not include in the announcement when the system becomes available to the public or any possible restrictions it may have. But they have a sign-up form for those who want to test the Make-A-Video.
This photograph taken on October 28, 2021, shows the META logo on a laptop screen in Moscow as Facebook chief Mark Zuckerberg announced the parent company's name is being changed to "Meta" to represent a future beyond its troubled social network.
How Does Make-A-Video AI System Works?
Make-A-Video AI system lets people turn text prompts into high-quality video clips built on Meta AI's recent progress in generative technology research and has the potential to give creators and artists new opportunities. The AI learns how the world looks from paired text-to-image data and how it moves from unlabeled video footage.
Meta promises that this AI system lets users bring their imagination to life by generating whimsical and one-of-a-kind videos using only short phrases or lines of text. Meta AI has shared further details in a research paper appearing in the preprint server arXiv and will release a demo experience soon.
Generative AI research pushes creative expression forward by letting users quickly and easily create new content with just a few words, bringing imagination to life. The introduction of the Make-A-Video AI system follows the announcement of Make-A-Scene earlier this year.
The Make-A-Scene is a multimodal generative AI method that gives people more control over their AI-generated content. It lets users create photorealistic illustrations and storybook-quality art using only words or a line of text and freeform sketches.
AI systems, like the Make-A-Video, use publicly available datasets that add extra transparency to the research. More details are available in their research paper and openly shared with the community for feedback to refine and evolve their approach to the emerging technology.
Examples of Make-A-Video
Advances in AI to create videos on demand pose ethical concerns and Meta noted that all content generated but their new AI system contains a watermark to make sure that viewers will know the video was produced by the AI and is not a captured video. It might eventually become public, but the AI is still in the trial period to make sure it is safe and intentional.
One video clip Meta AI featured shows a fairly realistic-looking spaceship landing on Mars, according to MailOnline. Another video clip also shows an astronaut floating in space from different vantage points.
They also showed two other videos that are a bit surreal. One was "a ballerina performing a beautiful and difficult dance on the roof of a tall skyscraper," and another was of a blue unicorn "flying over a mystical land."
Meta AI noted that they are committed to developing responsible AI and taking the necessary steps to reduce the production of harmful, biased, or misleading content. Researchers wrote that Make-A-Video uses qualitative and quantitative measures to set the new state-off-the-art in text-to-video generation.
RELATED ARTICLE: OpenAI Introduces 'Outpainting' for DALL-E That Expands the Borders of Any Picture
Check Out More News and Information on Artificial Intelligence in Science Times
