nerfstudio-project/gsplat

Health v2 · maintenance only

Adoption not part of the score

5589 stars · 946 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

gsplat is an open-source Python library with CUDA-accelerated, differentiable rasterization of Gaussians, based on 3D Gaussian Splatting for real-time radiance field rendering. It offers faster, more memory-efficient kernels with features like multi-GPU support, sparse rendering, and various camera/sensor models.

Use cases

• render 3D gaussian splatting scenes in real time
• train radiance fields from photos with differentiable rasterization
• reconstruct 3D scenes from images using CUDA-accelerated splatting
• run multi-GPU distributed gaussian splatting training
• render with fisheye, LiDAR, or FTheta camera models
• speed up nerfstudio-style novel view synthesis

When to choose

• you need fast, memory-efficient gaussian splatting rasterization on NVIDIA GPUs
• you want a maintained, feature-rich alternative to the original 3DGS implementation
• you need differentiable rendering integrated into PyTorch training pipelines
• you require multi-GPU or sparse rendering support

When to avoid

• you have no CUDA-capable NVIDIA GPU
• you need a turnkey application rather than a Python library
• your project targets CPU-only rendering

Facets

library · maturity active

graphics image-processing machine-learning gpu-computing computer-vision graphics machine-learning gpu-computing python cross-platform gaussian-splatting cuda rasterization neural-rendering radiance-fields differentiable-rendering 3d-reconstruction gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="nerfstudio-project/gsplat")