Facebook wants to make you profile picture “come alive” using a “Harry Potter-style” technique. Researchers say the new images can display emotions and they will be known as “reactive profiles”.
Still pictures will be able to react and move depending on how you feel about any updates you interact with.
The “Haha” reaction, for instance, will cause your profile picture to look like it’s laughing. The same rule applies to “Wow”, “Sad” and “Angry”.
The researchers, who first presented the technique at SIGGRAPH Asia conference held in Bangkok last month, also say that they believe that they can animate your whole head and upper body rather than just your face.
It is thought that the idea is to encourage people to interact – even if it is just to see their profile pictures come alive.
It is definitely a “work in progress” as some of the images are quite frightening.
“In this work, we are interested in animating faces in human portraits, and in particular controlling their expressions,” wrote Facebook researchers Johannes Kopf and Michael F. Cohen, and Hadar Averbuch-Elor and Daniel Cohen-Or from Tel-Aviv University, in a paper.
“As our results illustrate, our technique enables bringing a still portrait to life, making it seem as though the person is breathing, smiling, frowning, or for that matter any other animation that one wants to drive with. We apply our technique on highly varying facial images, including internet selfies, old portraits and facial avatars.
“Additionally, we demonstrate our results in the context of reactive profiles – a novel application which resembles the moving portraits from Harry Potter’s magical world, where people in photographs move, wave, etc.”
The researchers say the expressions come from “driving videos” of a different person’s face, and that they can be transferred to pictures of anyone else’s face using “2D warps that imitate the facial transformations”.
Their technique also adds “fine-scale dynamic details”, such as creases and wrinkles, and “hallucinates” areas that may be hidden, such as the inner mouth.
“We built on the fact that there is a significant commonality in the way humans ‘warp’ their faces to make an expression. Thus, transferring local warps between aligned faces succeeds in hallucinating facial expressions,” the researchers say.
“As we have shown, the internal features of the mouth of the driving video are transplanted to the target face to compensate for the disoccluded region.
“Although this transplanting violates our goal of keeping the identity of the target face features, the effect of this violation is only secondary, since humans in general are not as sensitive to teeth as a recognition cue.”