3D Gaussian Splatting, first presented at SIGGRAPH 2023, is rapidly gaining attention across industries such as real estate, construction, manufacturing, and autonomous driving. As of 2026, major companies including NVIDIA are accelerating their research and implementation, and the technology is emerging as a de facto standard. Continuing from articles published in March and April, XR Expert Yoshinaga from Avaxia Asia offers a practical guide to the basic workflow for running the technology with a minimal setup.
This is Yoshinaga from the Data Integration and Development Division.
Gaussian Splatting is a technique for reconstructing relatively wide outdoor scenes in 3D, and it has been gaining significant attention recently.
In this article, we introduce the minimum workflow to get started with Gaussian Splatting, aimed at those who want to give it a try.
Most of the tools featured here are free to use (some optional features are paid), so feel free to jump right in.
Example Output
The following is one example of a working configuration.
| Item | Example |
|---|---|
| OS | Windows 11 |
| GPU | NVIDIA GeForce RTX 4070 SUPER |
| CPU | AMD Ryzen7 8700G |
| RAM | 32 GB |
Any of the following 360° cameras can be used. 8K or higher resolution is recommended for best results.
| Software | Purpose |
|---|---|
| LichtFeld Studio v0.5.1 | GUI tool for 3D Gaussian Splatting |
| 360° Gaussian v1.4.0 | Tool that automates each step of the Gaussian Splatting pipeline |
To build it yourself, follow the build instructions here.
Tip 1
Also note: although not mentioned in the official build instructions, Perl may be required. Install it from Strawberry Perl.
Be aware that the cmake bundled with Perl may take precedence over your system cmake, so first temporarily rename C:\Strawberry\c\bin\cmake.exe, and delete it only if necessary. To be safe, restart your PC afterward.
Tip 2
In the official documentation’s Clone repository section, the following command did not always download all required files correctly in the author’s environment:
git clone https://github.com/MrNeRF/LichtFeld-Studio
Instead, to avoid missing submodule downloads, it is recommended to add the –recursive flag, or use a client tool such as Fork or GitHub Desktop:
git clone –recursive https://github.com/MrNeRF/LichtFeld-Studio
Tip 3
Note: skip the git checkout v0.4.0 command in the Checkout stable version step.
The build process takes a long time, so plan to run it when you have plenty of time available.
Gaussian Splatting generally involves the following 4 steps:
| # | Step | Description |
|---|---|---|
| 1 | Capture | Record the scene with a 360° camera |
| 2 | SfM (Structure from Motion) | Estimate the position of each image |
| 3 | Point Cloud Generation | Generate a point cloud from the estimated camera positions |
| 4 | Gaussian Splatting | Generate a 3D Gaussian Splatting model from the point cloud |
Just as with normal shooting, record the scene with your 360° camera. To keep processing time manageable, it is recommended to start with a short video under 1 minute while you get familiar with the workflow.
Export the recorded footage in Equirectangular format using the PC application provided by your camera manufacturer.
In this step, we use 360° Gaussian to estimate camera positions and generate a point cloud.
This step automatically masks and excludes areas not needed for Gaussian Splatting (such as people or the sky). While not mandatory, applying masks can improve camera position estimation accuracy and the overall quality of the 3D model.
If masking is not needed, you can skip to 2.3 SfM Configuration & Execution.
360° Gaussian allows you to choose and combine multiple tools for both SfM and Gaussian Splatting training. Here we introduce our recommended settings when using the free tools SphereSfM and LichtFeld Studio.
Verify that the camera positions and point cloud have been generated correctly.
Verify that the point cloud and images have been loaded correctly.
If you do not need to display the camera images, uncheck Camera Frustum in the Rendering tab on the right side of the screen.
Optional Settings:
Only apply the following if you created mask images with AutoMasker.
To save progress partway through, click Save Checkpoint to record the current state.
Below are links to 3D Gaussian Splatting models actually generated using the workflow described in this article. You can interact with them directly in your browser via SuperSplat.
In this article, we introduced the basic end-to-end workflow for generating 3D Gaussian Splatting from 360° images. By using this workflow as a starting point and adjusting the settings, you can apply it to a wide variety of scenes.
For more advanced settings and techniques to further improve quality, please refer to the following articles:
Feel free to experiment with different settings and find what works best for your use case.
