If there is a major evolution in design in recent years, it is the use of AI Whether in the creation ex nihilo of illustrations or even in the field of deepfake – within the framework of the use of faces of actors having undergone the marks of time – the simple fact of automating or at least of facilitating this kind of task allows an astonishing flexibility.
That’s the idea behind Sony’s (Sony Interactive Entertainment LLC) recently published patent, centered on a method of capturing facial animations generated through AI. Without going into too much technical detail, the basic principle revolves around filming an actor’s face with 3D cameras in order to take into account its depth, then recovering the whole thing in the form of a mesh of polygons. Source that will then be used by the animators to design the various expressions. It is up to them to fill in the gaps, to put everything in order, in order to stick as closely as possible to reality.
Here, Sony’s patented technique relies on the same initial process, but uses both 2D and 3D cameras with the aim of training an AI to capture facial expressions as efficiently as possible, minimizing the need retouching, which removes a good deal of the painstaking work from the animators. A source not only of saving time, but also potentially of money. Nothing to even question the place of the said animators – at least for the moment – technique more perceived as a real support. And it is on this precise point that Sony could press to distribute this standard and modify a good part of the field.