Meta debuts an artificial intelligence system that generates videos from just TEXT prompts – with some spooky, surreal results
- Meta’s new artificial intelligence can generate video clips from text only
- The AI system, called Make-A-Video, was trained using a combination of existing labeled images and unlabeled videos
- So far, it’s delivering results that are impressive but often come across as spooky or surreal
- Some of the video clips include a convincing landing of a spaceship on Mars, a ballerina dancing on top of a skyscraper, and an astronaut floating in space
advertisement
Meta introduced a new breed of art-generating AI that can create videos from nothing more than text prompts – with results that are spooky, surreal, and awe-inspiring.
The artificial intelligence system announced Thursday, called Make-A-Video, uses existing images with captions to learn more about the world and how it is described, and uses uncaptioned video to determine how the world is moving.
The resulting videos range from surreal to stylized and spooky to compelling.
Mark Zuckerberg’s company hasn’t revealed when the system will open to the public or if there will be any restrictions — but there is a sign-up form people can fill out if they want to try Make-A-Video in the future.
Meta introduced a new breed of art-generating AI that can create videos from nothing more than text prompts – with results that are spooky, surreal, and awe-inspiring
The artificial intelligence system announced Thursday, called Make-A-Video, uses existing images with captions to learn more about the world and how it is described, and uses uncaptioned video to determine how the world is moving
Advances in using AI to create on-demand videos raise all sorts of ethical dilemmas — not to mention the possibility of deepfakes and disinformation.
Meta noted that all content generated by the new system will include a watermark to ensure “viewers know the video was created with AI and is not recorded video.”
“Our goal is to eventually bring this technology to the public, but for now we will continue to analyze, test and test Make-A-Video to ensure that every step of the release is safe and intended,” the tech giant explains.
A video clip shows a fairly realistic looking spacecraft landing on Mars.
Another shows an astronaut floating in space from different angles.
Mark Zuckerberg’s company hasn’t revealed when the system will open to the public or if there will be any restrictions — but there is a sign-up form people can fill out if they want to try Make-A-Video in the future
Meta noted that all content generated by the new system includes a watermark to ensure viewers know the video was created with AI and is not recorded video.
Mark Zuckerberg’s Meta pays out billions of dollars to create virtual worlds where people shop, work, and entertain
Two other videos look a bit more surreal.
One shows a ballerina performing “a beautiful and difficult dance on the roof of a very tall skyscraper,” with an illuminated cityscape behind her.
Another shows a blue unicorn “flying over a mystical land.”
“Meta AI is committed to developing responsible AI and ensuring the safe use of this cutting-edge video technology,” the company wrote.
“Our research takes the following steps to reduce the creation of harmful, biased, or misleading content.”
“In all aspects, spatial and temporal resolution, text fidelity and quality, Make-A-Video sets the new state of the art in text-to-video generation, which is determined by both qualitative and quantitative measures,” the researchers write in a paper explaining the system.
Previous AI systems that generate still images from text have produced strange results.
Craiyon AI, another popular non-Meta text-to-image system, created several different images of what humans might look like if all humans joined the metaverse.
Everyone has an augmented reality headset fused to their face.
These images were first created when The US Sun asked the AI these questions about humans in the metaverse.
A number of technology companies, including Zuckerberg’s, are investing tens of billions of dollars into creating virtual worlds where people can shop, work, and be entertained.
The images come after another AI working in the same way produced bizarre images of the “last selfie ever taken,” which shows apocalyptic scenes of people in front of nuclear explosions, with rotting flesh and total devastation beyond.
Craiyon AI, another popular non-Meta text-to-image system, created several different images of what humans might look like if all humans joined the metaverse. Everyone has an augmented reality headset fused to their face