Nvidia has unveiled Neuralangelo, an artificial intelligence model capable of transforming 2D video clips into intricate 3D structures. This development promises enhanced usability for professionals across various fields, including art, video game development, and industrial digital twins.

Built on neural networks, Neuralangelo can translate complex textures—from glass panes to roof shingles—from 2D videos into 3D assets, says Nvidia. The resulting high-fidelity 3D reconstructions offer significant advantages for creators seeking to generate virtual objects rapidly for their projects, employing simple footage captured via smartphones.

Neuralangelo uses instant neural graphics primitives to capture finer details, effectively handling challenges like repetitive texture patterns, homogeneous colors, and strong color variations, according to Nvidia.

Neuralangelo functions by selecting various frames from a 2D video of an object or scene, shot from different angles. This process mimics an artist considering a subject from multiple perspectives to understand depth, size, and shape. The AI model then generates a preliminary 3D representation of the scene. Lastly, the model optimizes the render to enhance details, leading to a 3D object or large-scale scene ready for use in multiple applications.

Find the blog post from Nvidia on Neuralangelo here.

[Image courtesy: Nvidia]