CN/EN
CN/EN
Artificial intelligence (AI) is a fast-growing technology that is increasingly being adopted in many aspects of our daily lives. AI algorithms are used in a number of sectors, including medical, automotive and 3D printing. In this context, NVIDIA recently unveiled a new generative AI model called LATTE3D. It is capable of transforming text descriptions into 3D models in a matter of seconds (faster than you can brew a coffee according to the company!), similar to a virtual 3D printer. This technology can be used in various fields such as game development, design, robotics or any other application requiring a virtual environment.
Unlike conventional approaches, which require a time-consuming investment in the creation of 3D models, whether from scratch or by selecting from a library of existing objects, LATTE3D offers an alternative. This software is capable of generating models virtually instantaneously from text descriptions. For example, by providing a description of the desired object, LATTE3D can produce several corresponding 3D shape options in a matter of seconds. Thanks to its speed and precision, LATTE3D opens up new possibilities for designers.
Custom 3D Models From LATTE3D in Just a Few Seconds
LATTE3D allows users to choose from several 3D shape options to suit their needs. What’s more, selected objects can be optimized for best quality in just a few minutes. Once the model has been created, users can export it to other compatible graphics software or platforms.
“A year ago, it took an hour for AI models to generate 3D visuals of this quality — and the current state of the art is now around 10 to 12 seconds. We can now produce results an order of magnitude faster, putting near-real-time text-to-3D generation within reach for creators across industries,” explains Sanja Fidler, Sanja Fidler, vice president of AI research at NVIDIA.
While researchers refine LATTE3D by training it on two datasets, one for animals and the other for everyday objects, developers could exploit this same model structure to train AI on other types of data. For example, with a 3D plant dataset, LATTE3D could help a landscape architect create garden models. Similarly, using data on household objects, the model could generate elements for 3D house simulations.
In addition to this, the model was enriched by learning to interpret various textual formulations to describe 3D objects, with the help of ChatGPT. The model has therefore been enhanced by learning to understand and interpret different ways in which users can describe 3D objects using text. It can recognize a wider range of descriptions and instructions, making it more versatile and accurate in its application. It should be noted that LATTE3D is still under development and has not yet been commercialized. Researchers plan to extend it to other fields, such as 3D animation generation.
Join us at TCT Asia, connect with industry innovators as you explore the entire AM ecosystem including design, materials, hardware, software, post-processing, and quality. 10,000+ professionals will unite to hear about the latest trends, explore the latest immersive AM technologies and find solutions to their AM challenges. Make sure you are there too.
TCT ASIA 2024
Tuesday 7th May 09:00 - 17:30
Wednesday 8th May 09:00 - 17:30
Thursday 9th May 09:00 - 15:00
NECC(Shanghai)7.1&8.1H
