Pixal3D transforms any 2D image into a detailed, pixel-aligned 3D asset in seconds. It explicitly projects image pixels into a 3D volume, eliminating the spatial guesswork of canonical-space generators, and produces geometry with PBR materials (base color, normal, roughness) that closely match the input view. Developed by Tsinghua University and TencentARC researchers and accepted at SIGGRAPH 2026, it supports single or multi-view input and exports GLB meshes. Fully open source under the MIT license.