Chinese researchers are introducing a new technique that can transform photos of people into moving animations. Except for a few details, the videos appear lifelike.

With today’s AI tools, creating realistic photos and videos is no longer so difficult. AI is even getting better at depicting human figures. Until now, converting photos into moving images was still a difficult hurdle. Chinese researchers have developed a tool called Encourage everyone achieve a breakthrough.

To go from still images to moving images, an AI model distills physical properties from a person’s photo. Based on these features, a series of images of the same person in a different pose is then created using robust diffusion techniques. Inserting these images one after the other creates the effect of an animation.

This change is often difficult. The moving images have inconsistencies in the character’s physical appearance or clothing. In an article, Chinese researchers associated with Alibaba reveal how they solved this problem. The model takes an additional intermediate step to better understand the relationship between the reference image and a space, resulting in a better representation of the external features.

Animated dances

In a number of examples, we see Animate Everyone create both real people and 2D animations, and even humanoids perform a dance based on a single still image. The moving images look lifelike, but if you look closely you can still see some imperfections. The researchers point out that hands and eyes still represent the biggest obstacle.

Of course, there are ethical questions about this technology. The model is currently only intended for academic use, but you may wonder what happens when this technology or an equivalent variant is made available to the general public. Deepfakes are already causing problems for the credibility of visual content. As technology improves, people with bad intentions need nothing more than a photo from a social media profile to get you to do something.