Nerfstudio: 3D generator — what it does and what you need to run it

Nerfstudio is an open-source platform for creating 3D scenes using Neural Radiance Fields (NeRF) and related techniques such as Gaussian Splatting. The tool is aimed at researchers and developers who need not just a "button" but an environment where they can assemble a reconstruction pipeline from ready-made modules, swap components, and compare results.
What it does
- Reconstructs 3D geometry and scene appearance from a set of photos or video — the authors call it a "collaboration friendly studio for NeRFs".
- Supports several reconstruction methods: classic NeRF models and Gaussian Splatting (as indicated by the repository tags — nerf, gaussian-splatting, 3d-reconstruction).
- Allows you to train a model, visualize the result, and export geometry (mesh) — the authors do not specify the exact list of export formats in the fact sheet.
- The code is written in Python using PyTorch, which makes it possible to integrate Nerfstudio into existing machine learning pipelines.
- The project is accompanied by documentation on a separate site: docs.nerf.studio.
What you need to run it
- NVIDIA GPU with CUDA. The developer states it directly: "You must have an NVIDIA video card with CUDA installed on the system". The library was tested with CUDA version 11.8.
- Operating system — the authors do not specify hard restrictions, but the CUDA requirement implies Windows or Linux.
- The amount of VRAM and system RAM is not stated in the fact sheet; it can be assumed that the requirements depend on the resolution of the images and the chosen method.
- Installation is done from the repository: github.com/nerfstudio-project/nerfstudio.
- Code license: Apache-2.0. This permits commercial use and modification.
- This is a local tool, not a cloud service — you cannot run it without your own hardware with an NVIDIA GPU.
Who it is for
The tool is designed for computer vision specialists, researchers, and engineers who understand how NeRF and Gaussian Splatting work and are willing to configure the pipeline for their task. It suits those who need experiment reproducibility and the ability to change system components.
It is not suitable for artists or designers looking for a ready-made application with a "upload photos — get a model" interface. It is also not suitable for users without an NVIDIA GPU — on AMD or Intel, the tool, as stated by the developer, will not work.
Nerfstudio is an open-source research framework that gives full control over the reconstruction process but requires an understanding of the internal mechanics of the methods and compatible hardware.
Fact sheet
| Code license | Apache-2.0 source |
|---|---|
| Platform | CUDA per the author’s description source |
| Language | Python source |
| Last code change | 2025-07-29 source |
| Repository created | 2022-05-31 source |
| Latest release | v1.1.5 source |
|---|---|
| Release date | 2024-11-11 source |
| GitHub stars | 11871 source |
| Forks | 1657 source |
Values are collected automatically from official sources and were checked on 2026-08-19. Each one links to its source, and values taken from the developer’s pages also carry a verbatim quote — hover over the note. Pricing and versions are shown as of the check date and change most often; verify on the vendor’s site before buying.



