Microsoft has released a compact 4-billion parameter AI model that can convert any 2D image into a full 3D asset in just three seconds. The efficiency of the model is remarkable given the complexity of 3D reconstruction from a single image.

Traditional approaches to image-to-3D require significant computational resources and often take minutes to produce acceptable results. Microsoft’s model achieves near-instant results while maintaining high geometric and textural quality, making it practical for real-time applications.

The model’s small size (4B parameters) means it can run on relatively modest hardware, opening up 3D asset generation to game developers, AR/VR creators, and e-commerce platforms that need quick 3D previews from product photos.

Source