Skip to content

CRM

Turn a single image into a textured 3D mesh

Visit official site

What CRM is

The model is described as a high-fidelity convolutional reconstruction model that generates a 3D textured mesh from a single image. It integrates geometry priors into the network design and completes the generation within 10 seconds without test-time optimization.

What you can do with it

  • Generate a 3D textured mesh from a single image
  • Generate six orthographic view images
  • Build a high-resolution triplane with a convolutional U-Net
  • Use Flexicubes for end-to-end optimization

Application scenarios

  • Generate a 3D model quickly from a product photo
  • Create textured interactive 3D objects in virtual reality
  • Carry out 3D digitisation of cultural heritage

Model notes

  • Geometry priors are integrated into the network design
  • Six orthographic view images come from one input image
  • A convolutional U-Net creates the high-resolution triplane
  • The 10-second generation needs no test-time optimization

Pros & cons

✓ What we like

  • One image produces a high-fidelity textured mesh
  • Flexicubes supports direct end-to-end optimization on the textured mesh

! What to watch out for

  • The model takes a single image as its input
  • The 10-second figure is stated for producing a textured mesh from an image

FAQ

What is the input?

The material states a single input image is turned into a 3D textured mesh.

How fast is generation?

The material states the whole model produces a high-fidelity textured mesh from an image within 10 seconds, with no test-time optimization.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review