Training Framework

Nerfstudio

Nerfstudio is a modular radiance-field framework that includes Splatfacto, its Gaussian Splatting method, plus dataset tools, viewers, and a broad research ecosystem.

Stage: TrainingExperimentationDataset toolingNeRF and 3DGS teams

Open official linkNerfstudio Documentation->

What It Does

Nerfstudio is useful when your work crosses between NeRFs, Gaussian Splatting, research methods, and dataset preparation. Splatfacto is the 3DGS path inside that ecosystem.

It is more structured than standalone scripts but more technical than a desktop app. That makes it a strong default for reproducible experiments and custom pipelines.

How To Use It In 3DGS

Install Nerfstudio and verify GPU support.
Process images or video into a supported dataset format, often using COLMAP under the hood.
Train with Splatfacto or another relevant method config.
Use Nerfstudio visualization tools during training, then export for browser viewing or conversion.

Things To Watch

It is a framework, so expect environment and configuration work.
Method names and defaults can change as the ecosystem evolves; pin versions for papers or client projects.
Export expectations vary between viewers, so validate the final format before delivery.

Official And Related Links

Nerfstudio Documentation Splatfacto Documentation Nerfstudio GitHub gsplat Documentation

Splatfacto Notes

Nerfstudio calls its Gaussian Splatting implementation Splatfacto to distinguish it from the original INRIA paper code and to leave room for framework-specific improvements.
The strongest reason to choose Nerfstudio is the surrounding ecosystem: dataparser conventions, viewer tooling, method configs, and a large set of NeRF/3DGS-adjacent methods.
For custom research, combine Nerfstudio’s pipeline ergonomics with gsplat’s lower-level rasterization APIs instead of starting from a blank training script.

Splatfacto Docs Nerfstudio Methods