The power of artificial intelligence (AI) is starting to materialize now that we see real-world applications of the technology. Here’s another one and it’s from Google, the Imagen Video AII.
As the title states, the Google Imagen Video AI can produce videos using text. What’s more, Google’s AI technology combined image, video, and text datasets and was trained to create artistic videos.
- 13 ways how Artificial Intelligence will transform the future
- Free AI-powered image restoration tool fixes old photos in a click
While the initial results don’t look too impressive, it does prove that there’s a potential to use AI in the creative industry.
To train its Imagen Video AI, Google used an internal dataset that consists of up to 14 million videos and 60 million images. Google also used 400 million images via the LAION-400M open data set.
Thanks to this heavy dataset training, the tool can create a five-second video clip based on the supplied inputs.
We’ve seen similar technology done on images. But unlike those, you have to be more descriptive when telling the tool how you want the video would be.
Google Images Video AI works like DALL-E 2, a similar technology released earlier by a smaller company. It may have better results than Imagen Video, but that’s because Google’s tech has to deal with more complex things like 3D objects and how they interact with each other.
That said, being a giant company with a lot of money for research and development, we expect the Google Imagen Video to become a whole lot better in the future.