NVIDIA introduces LATTE3D, a new generative AI model capable of transforming text inputs into complete 3D models within seconds.
LATTE3D revolutionizes 3D modeling with instant results and offers applications in various fields such as game development, design, and robotics.
The technology promises to accelerate workflows and significantly enhance efficiency in 3D modeling.
NVIDIA Unveils LATTE3D: The Next Chapter in 3D Printing Enables Instant Conversion of Text to 3D Models.
The AI research team at NVIDIA has introduced a new generative model capable of transforming text inputs into 3D models within a second, making it ready for applications such as game development, design, and robotics (or at least on the verge). So far, the model has only been trained for generating "animals and everyday objects," but additional categories can be added. The nearly instant speed is expected to change many workflows.
While AI models for image and video generation have dominated the media for some time, NVIDIA has taken the next logical step with LATTE3D, a new generative AI model that converts text inputs into complete 3D models in less than a second.
The technology, announced at the NVIDIA GTC last week, is touted as a kind of "virtual 3D printer" that delivers instant 3D models "in a popular format used for standard rendering applications," for use in game development, advertising, design projects, or any application requiring a virtual environment, including VR and AR.
Generative text-to-3D AI is not a new technology, and NVIDIA's research paper itself compares LATTE3D to several existing systems such as MVDream, 3DTopia, and LGM. The overwhelming advantage of LATTE3D is its speed. In comparison tests, other AI models require between 20 seconds and 30 minutes to generate an output, while LATTE3D delivers a comparable (and often better) output in just 400 milliseconds.
Like with 3D printing, the quality of LATTE3D and other AI models' output improves over time. The advantage is that, unlike with 3D printing, you can initially prioritize speed over quality and later wait for a refined output instead of setting parameters from the beginning.
Like any other generative AI, LATTE3D has the potential to change many workflows. While previously it was necessary to search through a 3D asset library or spend valuable hours creating a simple object, with NVIDIA's technology, it suffices to input a brief description and import the generated output into one's own software application or platform.
