NVIDIA has just presented an AI model to create objects and characters in 3D: a new tool to build the metaverse

NVIDIA has just presented an AI model to create objects and characters in 3D: a new tool to build the metaverse
nvidia has just presented an ai model to create objects

Building the metaverse, the omniverse, and the next generation of video games will require the creation of an enormous number of virtual objects. In response to this need, NVIDIA has introduced a artificial intelligence (AI) model spawning a diverse array of buildings, vehicles, and 3D characters.

As the company explains, the model, called NVIDIA GET3D, has been trained in just two days with around a million 2D images of 3D shapes captured from different camera angles using its A100 GPUs. These are graphics processing units with 54 billion transistors.

A solution to facilitate the construction of the metaverse

Imagine having the possibility of creating three-dimensional objects “with high-fidelity textures and complex geometric details” in a simple way and prepared for subsequent adaptation to different types of projects. This is what NVIDIA proposes with GET3D.

For example, when building a virtual universe, you could create thousands of performances that will be part of the buildings, outdoor spaces, cars, roads, animals and characters. Each with its own distinctive features.

GET3D can generate images based on the data it is trained on. Thus, from 2D car images you can create a collection of other cars with different designs. The same goes for animals, you can create different versions of the same species.

One of the advantages offered by this model is that the shapes generated, thanks to their 3D mesh structurethey can be imported into game engines, movie rendering programs and 3D modellers to adapt them to different projects.

SEE ALSO  Can I connect to my neighbor's WiFi printer to play a prank on him?

All those interested in exploring this latest research work can do so by consulting the project’s GitHub repository. The company ensures that the code, under an open source license, will be available next week.

In Xataka | The Ada Lovelace Architecture of NVIDIA’s GeForce RTX 40 GPUs Explained