Alibaba’s Animate Anyone AI model turns pictures into videos. It was trained on TikTok videos

The research team of the Chinese company Alibaba has published a paper describing a new AI model called Animate Anyone. It is able to transform static images into short videos, writes 404 Media.

Several examples of how Animate Anyone works are already published on the X platform.

The videos created by the model look quite convincing. So some people assume that artificial intelligence may soon replace TikTok users who post “dancing” videos.

However, users also point out that Animate Anyone can be used to create fake content for malicious purposes.

At the same time, the publication draws attention to the fact that the model was trained using videos of TikTok users with a large audience. We are talking about dozens of authors. And all this could have happened without their consent.

Among them is the dancer and actress Maddie Ziegler, best known for her participation in the reality show Dance Moms. She has 23.5 million followers on TikTok. Another example is blogger Jasmine Chiswell, who has almost 17 million followers on this social network.

The company’s research team indicates that to train the model, it used a dataset from TikTok consisting of “340 training and 100 test videos” of a single person dancing. This dataset came from a project at the University of Minnesota.