Learn how to create a multi-camera dataset for Gaussian Splatting directly in Unreal Engine using Camera Render Studio. The Gaussian Shots workflow can render an object or scene from multiple virtual cameras and prepare structured image and camera data for Postshot, COLMAP, LichtFeld Studio, and other compatible Gaussian Splatting reconstruction workflows.
Unreal Engine → Camera Render Studio Gaussian Shots → Multi-Camera Rendering → Postshot / COLMAP Dataset → Gaussian Splat Reconstruction
Creating a Gaussian Splatting Dataset in Unreal Engine
Camera Render Studio includes a dedicated Gaussian Shots workflow for generating synthetic multi-view datasets directly from an Unreal Engine scene.
The process requires only a few steps: create a Gaussian capture rig, select the camera distribution, configure the Movie Render Queue export, render the dataset, and then use the resulting images and camera information in your preferred Gaussian Splatting reconstruction application.
Step 1 — Add GaussianCaptureRigActor
Open the Camera Render Studio menu in the Unreal Engine toolbar.
Select: Camera Render Studio -> Add Gaussian Capture Rig Actor
Camera Render Studio will automatically create a GaussianCaptureRigActor in the current Unreal Engine level.
The GaussianCaptureRigActor controls the virtual camera layout used to capture the object or scene from multiple viewpoints.
Step 2 — Position the Gaussian Capture Rig
Move the GaussianCaptureRigActor to the object or area that you want to capture.
For an individual object or character, place the GaussianCaptureRigActor approximately in the center of the object.
The virtual capture cameras will be generated around this position according to the selected camera layout.
For environments or larger scene captures, the rig can instead be configured to distribute cameras throughout a defined volume.
Step 3 — Choose the Capture Layout
With the GaussianCaptureRigActor selected, open the Capture Rig settings and choose the required Layout.
Camera Render Studio provides several layouts for Gaussian Splatting capture:
- Spherical (Object Capture)
- Cylindrical (Object Capture)
- Volume Grid (Scene Capture)
- Volume Grid Spline (Scene Capture)
For individual objects and characters, Spherical or Cylindrical capture is generally the most suitable starting point.
For larger environments and scenes where cameras need to be positioned throughout an area, use Volume Grid or Volume Grid Spline.
GaussianCaptureRigActor → Spherical, Cylindrical, Volume Grid and Volume Grid Spline capture modes
Step 4 — Configure the Camera Count
Choose the required Camera Count and configure the capture parameters for the selected layout.
Depending on the selected capture mode, Camera Render Studio allows you to configure parameters such as:
- Camera Count
- Radius
- Horizontal FOV
- Minimum Elevation
- Maximum Elevation
- Camera placement and distribution
A larger number of cameras provides more source viewpoints for reconstruction, but it also increases rendering time, dataset size, storage requirements, and Gaussian Splat reconstruction time.
Tip: Choose the camera count according to the complexity and size of the captured object or scene. Camera Render Studio allows you to preview the generated camera positions directly inside the Unreal Engine viewport before rendering.
Step 5 — Add Postshot / COLMAP Dataset Export
Open Movie Render Queue and configure your render job.
Add the dataset export module:
+ Setting → Exports → Postshot / COLMAP Dataset
In Camera Rig Override, select the GaussianCaptureRigActor that was created in the level.
Then choose the appropriate dataset format and output settings for your reconstruction workflow.
Step 6 — Configure the Dataset Export
The Postshot / COLMAP Dataset module prepares the camera information required by compatible Gaussian Splatting reconstruction applications.
Configure the required settings:
- Camera Rig Override
- Use MRQ Output Directory
- Export Directory
- Dataset Format
- Frame Mode
- Frame Number, when required
For a static object, you can use Single Frame.
Camera Render Studio exports camera information used by compatible multi-view reconstruction workflows, including camera position, rotation, field of view, and intrinsic camera parameters.
Step 7 — Add Camera Render Studio (GaussianShot)
In Movie Render Queue, add the Camera Render Studio Gaussian rendering module:
+ Setting → Rendering → Camera Render Studio (GaussianShot)
In the Camera Source section, select the same GaussianCaptureRigActor.
This module controls how Camera Render Studio processes and renders the virtual cameras generated by the Gaussian capture rig.
Movie Render Queue → Camera Render Studio (GaussianShot)
Step 8 — Split Large Camera Sets into MRQ Tasks
For Gaussian Splatting datasets containing many virtual cameras, it is recommended to use:
Render Mode → Split Into MRQ Tasks
This allows Camera Render Studio to divide a large camera set into several smaller Movie Render Queue tasks instead of processing all cameras as one large render task.
Set the required Cameras Per Task value.
Then click:
Create Camera Tasks
Camera Render Studio will automatically calculate the number of required tasks and create them for the render.
Recommended: Use MRQ task splitting when working with large Gaussian capture rigs. Dividing the cameras into multiple render tasks helps reduce peak memory usage and makes large datasets easier to render and manage.
Step 9 — Configure Object Masks
Camera Render Studio can also generate additional masks for Gaussian Splatting workflows.
Under Post Shot Masks, choose the required Actor Layer Mask Mode.
Available modes include:
- Hair / Full Object (ShowOnlyPrimitives)
- Standard Stencil (Preserve Occlusion)
Hair / Full Object
The Hair / Full Object (ShowOnlyPrimitives) mode can be useful when working with characters and complex hair geometry.
Standard Stencil
The Standard Stencil (Preserve Occlusion) mode uses the standard stencil-based workflow while preserving occlusion.
You can also configure Actor Layers and Data Layers according to the requirements of the scene.
Actor Layer Mask Mode → Hair / Full Object or Standard Stencil
Step 10 — Start Rendering
Once the Gaussian capture rig, camera layout, dataset export, GaussianShot settings, MRQ tasks, and optional masks are configured, start the render from Movie Render Queue.
Camera Render Studio will process the virtual cameras generated by the GaussianCaptureRigActor and render the required image sequence while preparing the associated camera dataset.
If MRQ task splitting is enabled, the camera set will be distributed across the generated Movie Render Queue tasks.
Step 11 — Create Your Gaussian Splat
After rendering is complete, the output directory will contain the source material required for Gaussian Splat reconstruction.
The generated images and camera data can then be used with compatible Gaussian Splatting reconstruction workflows such as:
- Postshot
- COLMAP-based Gaussian Splatting workflows
- LichtFeld Studio
- Other compatible Gaussian Splatting reconstruction tools
The final Gaussian Splat is reconstructed by the selected external application using the source images and camera information generated by Camera Render Studio.
Interactive Gaussian Splat Example
The interactive example below was created from a MetaHuman rendered directly in Unreal Engine using Camera Render Studio. Camera Render Studio generated the multi-camera source images and camera dataset used for the Gaussian Splatting reconstruction workflow.
Interactive MetaHuman Gaussian Splat generated from source images rendered with Camera Render Studio for Unreal Engine. Open in SuperSplat
Why Use Camera Render Studio for Gaussian Splatting?
Creating synthetic Gaussian Splatting datasets manually can require placing and managing a large number of cameras, rendering every viewpoint, tracking camera transforms, exporting camera parameters, and organizing the resulting files.
Camera Render Studio Gaussian Shots automates this workflow directly inside Unreal Engine.
- Generate structured multi-camera capture rigs.
- Capture individual objects, characters, or complete environments.
- Use Spherical and Cylindrical object capture layouts.
- Use Volume Grid and Volume Grid Spline layouts for scene capture.
- Preview generated camera positions directly inside Unreal Engine.
- Configure camera count, field of view, radius, elevation, and camera placement.
- Automatically render large numbers of virtual cameras.
- Split large camera sets into multiple Movie Render Queue tasks.
- Reduce peak memory usage during large multi-camera renders.
- Export camera transforms and intrinsic camera parameters.
- Generate Postshot / COLMAP datasets.
- Prepare datasets for LichtFeld Studio and compatible reconstruction workflows.
- Create optional masks for characters, hair, and other scene elements.
- Prepare static or frame-dependent multi-view datasets.
Gaussian Splatting Workflow
The complete Camera Render Studio workflow is:
Unreal Engine Scene → GaussianCaptureRigActor → Camera Render Studio (GaussianShot) → Movie Render Queue → Postshot / COLMAP Dataset → Gaussian Splat Reconstruction
Camera Render Studio does not perform the final Gaussian Splat reconstruction itself. Instead, it automates the Unreal Engine capture and dataset generation stage, providing the rendered multi-view images and camera information required by compatible Gaussian Splatting software.
Camera Render Studio for Unreal Engine
Camera Render Studio provides professional rendering workflows for Unreal Engine, including Gaussian Splatting dataset generation, VR180, 360 rendering, FullDome, projection mapping, multi-camera rendering, Movie Render Queue, and Movie Render Graph workflows.
Learn More About Camera Render Studio
