Nvidia Unveils AI Model for Generating 3D Objects from Prompts

Nvidia has unveiled Latte3D, a tool for instantaneously transforming text into 3D figures.

It can create high-quality three-dimensional images almost instantly.

“Latte3D was created by the Nvidia AI lab team in Toronto. It represents a significant advancement in AI, offering real-time creation of 3D objects from text prompts,” the press release states.

The model was trained using the A100 Tensor Core GPU.

According to Nvidia representatives, users can export the created figures to various applications, such as Omniverse, a software-hardware platform for creating metaverses.

In Latte3D, the company employed amortized methods. This mechanism allows the model to break down a text prompt into various components for rapid output. With this, the company aims to address the slow generation issues in earlier versions.

Using a single GPU, such as the Nvidia RTX A6000, the model can instantly create various three-dimensional objects without the labor-intensive rendering process.

Latte3D offers several options based on each input prompt, allowing users to select the most suitable version.

The AI model has proven to be much more versatile than anticipated. For instance, it can be applied in fields such as landscape design and robotics.

Previously, Nvidia announced a new generation of AI chips and software for working with large language models.

Back in March, the company’s CEO, Jensen Huang, introduced a multimodal AI developed specifically for androids.