Google’s DeepMind neural network now has an important capability. Thanks to a new tool called Trasframer, short videos can be created from a single frame of picture given to artificial intelligence.
Displayed using Google’s DeepMind artificial intelligence platform, the new technology Transframer can create short 30-second videos from a single image input. According to Google, all Transframer needs is just one photo.
What is DALL-E 2 and Midjourney?
The developed system firstly analyzes the photo used as input and identifies the image and its frame, that is, it guesses whether it is a desk or a computer. Afterwards, object and environment analysis is performed based on the context of the image. As a result of all these processes, short, three-dimensional videos are created. The videos produced create images that contain different perspectives of the object in context or the direct image, although there is no geometrical data.
Could be revolutionary for games
The DeepMind team tested Transframer on various objects such as chairs, laptops, glasses and books in a series of tests they carried out, and the result was extremely successful. Although the videos created seem consistent, of course, let’s underline that this technology is still in the early stages. However, predictions have already begun to be made about the future of Transframer. According to Futurism, Transframer could usher in a new era in the video game industry by using machine learning to create digital environments in the future.
Neural networks and artificial intelligence systems continue to develop rapidly in recent years. DALL-E 2, the machine learning model developed by OpenAI, produces similar visuals by only making sense of the words or sentences given to it, unlike Transframer. Some Twitter users even wonder how Transframer converts an image created in DALL-E 2 to video.
906
