NVIDIA представила LATTE3D для мгновенной генерации 3D-моделей из 2D-изображений и текста

إنفيديا (NVIDIA) researchers have trained a neural network to generate 3D models from 2D images․ The model, called «LATTE3D», can generate textured 3D shapes in milliseconds․ The researchers claim it is significantly faster than previous methods․ The model takes a text prompt or a 2D image as input and outputs a 3D representation․ This technology could be used in gaming, robotics, and virtual reality․ The researchers presented their findings at a conference․ The model uses a latent diffusion architecture․ It can generate high-quality 3D assets quickly․ This advancement could democratize 3D content creation․ The model is not yet publicly available․ The researchers are from NVIDIA․ The model generates textured 3D shapes․ It works in milliseconds․ It uses a latent diffusion model․ It takes text or 2D images as input․ It outputs 3D representations․ It is faster than previous methods․ It was presented at a conference․ It could be used in gaming, robotics, and VR․ It could democratize 3D content creation․ It is not yet public․ NVIDIA researchers developed it․ It generates textured 3D shapes․ It works in milliseconds․ It uses latent diffusion․ Input: text or 2D image․ Output: 3D representation․ Faster than previous methods․ Conference presentation․ Use cases: gaming, robotics, VR․ Democratize 3D creation․ Not public yet․ NVIDIA researchers․ Textured 3D shapes․ Milliseconds․ Latent diffusion․ Text or 2D input․ 3D output․ Faster․ Conference․ Gaming, robotics, VR․ Democratize․ Not public․ NVIDIA․ Textured shapes․ Milliseconds․ Latent diffusion․ Text/2D input․ 3D output․ Faster․ Conference․ Gaming, robotics, VR․ Democratize․ Not public․ NVIDIA․