3D AI Advancements: Pixel 3D, NVIDIA, and More

Summary

Open-source 3D AI is rapidly evolving, with Pixel 3D now available in a low VRAM GGUF version, requiring only six gigabytes of memory and released under an MIT license for commercial use. NVIDIA has introduced a real-time AI motion system for games called Motion Bricks, which generates character movements on the fly based on provided input, also open-sourced under an Apache 2.0 license. Helix 4D offers 4D mesh generation from video input, creating animated 3D meshes. WaiWo 3D is developing photorealistic world generation, with initial impressive examples. Apple's Littor for image-to-3D generation is now integrated into Comfier UI. NVIDIA also presented SANA WM for real-time video generation, though it requires high-end hardware like an RTX 5090 to run. Meta AI and Technion developed Realis 3D, a framework to improve textures on existing 3D models, with code expected soon. The Roblox team released a cube model for parts and assembly with a large labeled dataset. Rest 3D reconstructs physically stable 3D scenes from single images, preserving mass and interaction properties. Finally, AssetHub provides a comprehensive 3D AR pipeline tool, and updates to 3D Gaussian Splatting, including a hollow viewer capable of handling up to one billion splats, further enhance its capabilities.