The development of artificial intelligence (AI) technology is now accelerating and is seen to change many aspects of modern life such as threatening some careers. Recently the new AI module Animate Anyone was shown with it can produce an animated video by just providing a still image.
It was developed by Alibaba Group's Institute for Intelligent Computing which leverages the support of generation or generative technology. It can also be used for various objects including humans and cartoons with high-quality, clear and stable video results. Through testing, Alibaba was seen to focus on certain things including complex dance movements.
Technically there are three main components used in this technology including RefenrenceNet which is designed to combine detailed features while preserving details from the reference image. Followed by Pose Guider for more realistic character movement control in line with the desired style and Temporal Layer for smooth animation transitions in the video frame.
In the demo shown, video dance data on TikTok was used in training this model. Any dance on TikTok can be reproduced by Animate Anyone using any selected still image. Among its uses in the real world is to produce animations of models showing off clothes in online stores.
Currently there is no dedicated site that allows Animate Anyone to be used by all but there is a repository on Github for those interested in trying it out. This technology is considered interesting but must be done responsibly to avoid the spread of inappropriate content that can manipulate sentiment in society.