OpenAI launches Shap ·E: generating high-quality 3D models based on text in seconds

CTOnews.com May 9 news, following the text-generated picture model DALL E, OpenAI recently released the Shap E model again. Users can enter text to create realistic and diverse 3D models.

Shap ·E is not only a 3D model generator, but also can directly generate parameters of implicit functions (implicit functions), which can render texture meshes (textured meshes) and neural radiation fields (NeRF).

This means that Shap ·E, unlike current models that output only point clouds (point clouds) or voxels (voxels), can generate high-quality 3D assets with fine-grained textures and complex shapes.

OpenAI shows the results of Shap E. such as a bowl of food, a penguin, a voxel dog, a bonfire, an avocado-shaped chair, etc. The whole picture can be rendered in seconds.

CTOnews.com with related nouns is explained as follows:

Neural radiation field (NeRF): neural radiation field (Neural Radiance Field, NeRF), as a new field of view synthesis technology with implicit scene representation, has attracted wide attention in the field of computer vision.

As a novel method of view synthesis and 3D reconstruction, NeRF model is widely used in the fields of robot, city map, autonomous navigation, virtual reality / augmented reality and so on.

Point cloud (point clouds): a cloud point is a series of points that fall in a three-dimensional coordinate system that can be unfolded on the surface or facade of the building to capture the appearance of the building.

Voxels (voxels) are pixels in 3D space. A quantitative, fixed-size point cloud. Each unit is of fixed size and discrete coordinates.