Pixal3D is an image-to-3D AI model that converts a single 2D image into a high-fidelity 3D asset with detailed geometry and textures.
Context
-
tokens
Input
-
per 1M tokens
Output
-
per 1M tokens
Downloads
-
Pixal3D creates 3D images that align with pixels. It keeps the visual details, shapes, and angles of the original image better than old image-to-3D methods. Search results describe it as a model from Tencent ARC Lab and collaborators, focused on creating 3D assets from one or more images.
Faithful Image-to-3D Results
Pixal3D generates 3D models that closely preserve the original shape, style, silhouette, and visual identity of your input image, helping you create assets that look more consistent with the source reference.
Pixel-Aligned 3D Generation
With pixel-aligned 3D generation, Pixal3D builds a stronger connection between 2D image pixels and 3D space, producing clearer and more accurate image-to-3D results.
Detailed Geometry Reconstruction
Pixal3D captures object contours, structural details, and complex shapes, reducing the risk of overly simplified or generic 3D models.
PBR Texture Generation
Generate 3D models with rich physically based rendering textures, making your GLB previews and 3D asset projects look more realistic and production-ready.
Fast Model Generation
Pixal3D completes model generation in about 10 seconds, making it easier to test image-to-3D workflows, preview GLB results, and prepare STL-style 3D model drafts faster.
Step 1: Follow TRELLIS.2 Installation
Please first follow the installation guide of TRELLIS.2 to set up the base environment.
Step 2: Install Additional Dependencies
pip install -r requirements.txt
Step 3: Install natten
NATTEN_CUDA_ARCH="xx" NATTEN_N_WORKERS=xx pip install natten==0.21.0 --no-build-isolation
Please replace xx with the CUDA architecture and the number of build workers suitable for your machine.
Step 4: Install utils3d
pip install https://github.com/LDYang694/Storages/releases/download/20260430/utils3d-0.0.2-py3-none-any.whl
Note: requirements-hfdemo.txt is for the Hugging Face Spaces demo (H-series GPU architecture) and may not be compatible with other architectures.
