Transform Images into Editable PartPacker 3D Models with AI

PartPacker 3D is NVIDIA's revolutionary AI technology for efficient part-level 3D object generation. It uses Dual Volume Packing strategy and Diffusion Transformer architecture to convert single 2D images into editable PartPacker 3D models with separable components.

PartPacker 3D generates 3D models with editable, separable parts from just a single image input. Unlike traditional methods that create monolithic models, PartPacker 3D's Dual Volume Packing creates interconnected PartPacker 3D components that can be individually manipulated and animated.

Yes! PartPacker 3D models generated with our technology can be used for research, commercial applications including game development, animation, 3D printing, VR/AR, and interactive media. Check licensing terms for specific usage rights.

PartPacker 3D requires approximately 16GB GPU memory (float16) for inference. It's built on PyTorch 2.5.1 with CUDA 12.1 support. The PartPacker 3D web demo runs on NVIDIA's infrastructure, so you can use it directly through your browser.

PartPacker 3D accepts single 2D RGB images with 518×518 resolution recommended for optimal results. Supported formats include JPG, PNG, and WEBP. The AI analyzes the image structure to generate corresponding PartPacker 3D parts.

PartPacker 3D models can be exported in STL, 3MF, and other formats optimized for multi-material 3D printing and interactive applications. Each PartPacker 3D part can be separated and used independently in 3D software, game engines, or animation tools.