NVIDIA this month announced Magic3D, the company’s new artificial intelligence (AI) tool capable of creating three-dimensional (3D) object models from text descriptions. This novelty is especially useful for game developers, facilitating the process of rendering animals, vehicles and scenarios.
The new AI dispenses with modeling skills, managing to create the material around 40 minutes after the description has been entered, a time that can vary depending on the requirement of details. In one example, the following text was inserted “A poisonous blue frog sitting on a water lily” and the result was surprising — see below.
This is not the first time that a company has launched a platform that uses artificial intelligence to generate results, as even Google has a platform that transforms any text into photorealistic images. In the case of NVIDIA, the solution created is aimed at the gamer community.
According to the technicians responsible for the project, Magic3D comes as a response to Google’s DreamFusion, which generates three-dimensional results starting from a 2D image optimized in NeRF volumetric data, then the low-resolution raw file is optimized to create a quality 3D model. .
See below the result of the description of the frog:
Apparently, Magic3D’s processing power is twice as fast compared to DreamFusion, indicating that the NVIDIA version is better optimized. “We hope that with Magic3D we can democratize 3D synthesis and open up everyone’s creativity in creating 3D content,” said one of the researchers.
In summary, the tool should contribute positively not only to the creation of games, but also to expanding the virtual reality (VR) universe, which has been growing rapidly thanks to strong investments by Meta (formerly Facebook) in expanding the metaverse.