Neural Rendering is nothing more than the use of software techniques in hardware designed for artificial intelligence, specifically the branch of Deep Learning, for the improvement of computer generated graphics. This discipline takes advantage of the inclusion of so-called Tensor units in NVIDIA, AMD and Intel GPUs to accelerate the execution of a series of algorithms that improve image quality and graphic fidelity.
Because it is necessary?
GPUs do not create the images you see on the screen out of nowhere, but the graphic designers and animators spend long hours building the models that you will then see with your own eyes while you play. As the detail increases, the number of hours to build the different elements grows the workforce. All this leads to the need to have large offices and a multitude of personnel who spend long hours.
The consequences of it? A total stagnation of the visual level of the games which is beginning to be exemplified in the new generation of consoles and the fact that the catalog is full of intergenerational titles.
Well, one of the things that Neural Rendering does is to solve this general problem and it is important to emphasize this. Because if the video game industry for both PC and console cannot make higher quality graphics, it is not due to hardware limitations, but human. Therefore, the automatisms that allow to solve this bottleneck with the minimum possible amount of resources have to be sought and developed.
How does Neural Rendering improve the lives of developers?
We have prepared a few examples of how the combination of artificial intelligence and computer graphics they will make it easier for workers in video game studios and they will help to uncover the productive cap of the sector.
- Automatic detail generation and destruction of each element on screen according to the distance without graphic artists having to create several models of the same object.
- Automatic and procedural creation of fixed elements in the scene, such as trees, textures, the structure of the scenes, and so on.
- We also have the problem that representation of human persons in a credible way. To achieve greater fidelity, the so-called Subsurface Scattering is used together with long hours of work by the graphic artists in order to better simulate human skin. Neural Rendering can be trained to help generate models closer to reality
- One of the fields that has been developing for several years is the automatic facial animation which consists of the AI being able to animate a face and therefore move the vertices automatically according to a script or written text. This means a huge time saving in many contemporary games.
Are not all Neural Rendering applications, but they are an example of those that will make life and work easier for studios that create video games. It will not mean a destruction in jobs either due to the fact that there is a constant demand for games with more and more visual fidelity, but it must be borne in mind that these technologies serve much more than raising the resolution with the DLSS on duty. .
