Make-A-Video: This AI from Meta can create video from text
Text-to-visual generators are slowly growing in popularity. Until now, we've only had image synthesizers like Dall-E and Stable Diffusion. Meta has entered the field with an AI that can make videos from text or image prompts. It's called 'Make-A-Video.' This will help creators and artists to create content with ease. Meta also has a text-to-image generator called 'Make-A-Scene.'
Why does this story matter?
- AI text-to-image generators have been praised a lot for their ability to create great content. We expected text-to-video to take more time, but Meta surprised us with one.
- The technology is mind-blowing. But what does a more evolved version of Make-A-Video mean for human creators?
- Earlier this month, an AI-generated art won a competition. So, are humans slowly being pushed out of the race?
The AI can animate a static source image
With Make-A-Video, creators can use text or image prompts to create a video. It can also animate a static source image. This AI text-to-video generator is based on the existing text-to-image creators. To create the system, the company took images with descriptions and applied unlabeled video training data. With this data, the AI can predict what comes next after the image.
There are plenty of sample videos on the official website
On the Make-A-Video official page, Meta has shared some samples for us to see. There are examples of making a video from texts. There are also examples of creating videos from a static image or filling up the void between two images. In another section, Meta showcases how you can create variations of a video. You can visit https://makeavideo.studio/ to see the samples.
Meta plans to release a demo of the tool
Meta is yet to make any announcement regarding how and when Make-A-Video will be available to the public. The company said that it is planning to release a demo of the tool. There is a sign-up form on the official website for those who are interested in the tool. If you're interested, you can give image requests to Meta's AI team via Twitter.
There are concerns about AI tools being misused
As creating images and videos gets easier courtesy of artificial intelligence and machine learning, new concerns have risen. Many worry that these AI tools could be abused. On the official site, Meta says that it will apply filters and a watermark to prevent harmful content from being generated. The question is, how sustainable is that? With time, someone is bound to bypass these restrictions.