We’ve all seen DALL-E at this point, the image-generating AI that turns your requests into digital artworks. Google has taken this idea one step further with their latest creation, Imagen Video. The artificial intelligence-driven product will construct looping video clips based on user input in the form of words and phrases. Just as with DALL-E, the results of this software are equal parts bizarre and impressive.
This isn’t novel technology, though. Meta’s Make-A-Video predates Google’s software. Especially given that Google hasn’t released the software to the public yet, unlike Meta. If you’d like to try it, you can actually sign up for Make-A-Video right now. Google hasn’t left future customers completely in the dark, though. The company has cited issues with safeguarding as the reason why it is reluctant to release the software. However, it is likely that the software will eventually be made available to the public.
If the name Imagen sounds familiar to you, it should. This isn’t Google’s first foray into user-determined digital asset creation. Google unveiled a similar image generation tool to DALL-E in may of this year. However, while DALL-E prides itself on artwork-like images, Google’s Imagen tried to mirror photorealism wherever possible.
The new Imagen Video utility works in largely the same way. Google claims that the software features ‘a high degree of controllability and world knowledge.’ Therefore, Google is confident that the software can generate a wide array of high-fidelity videos in various aesthetic styles. The company says that the tool also has a deep understanding of 3-dimensional digital assets. This is what the company says leads to the tool being able to quickly create 24fps high-definition videos.
Now, going back to why the company says it doesn’t yet trust its creation. Google tries to filter undesirable content out as effectively as possible, but it admits that there is still a ‘risk that Imagen has encoded harmful stereotypes and representations.’
While Google is messing about with AI-driven image and video generation, the company is also trying to make the Google Search interface feel more natural.