Apps

Spatial video recording and playback, 3D Gaussian scenes, and desktop transfers for Arcturus Vision Camera.

Arcturus Vision Camera

Arcturus Vision Camera

Capture it. Relive it.

Record and view 3D spatial videos with Arcturus Vision Camera on Steam Frame.

  • Sample spatial videos included
  • Windowed and immersive viewing modes
View on Steam Steam Frame · Free
Arcturus Vision Gaussian XR

Arcturus Vision Gaussian XR

Step inside your memories.

Explore 3D scenes on Steam Frame, built with our tools or your favorite Gaussian splatting tools.

  • 8 example scenes included
  • View your own compatible PLY and SOG scenes
View on Steam Steam Frame · Free

Windows & Mac · Beta Ubuntu & Steam Machine (SteamOS)

Arcturus Vision Sync

Transfer recordings from Steam Frame and convert videos for other headsets and players on your computer.

  • Transfer selected recordings
  • Skip videos already downloaded
  • Convert on Windows, Mac, or Linux (Beta) for Apple Vision Pro, Meta Quest, standard 3D players, and 2D video.
Download for WindowsDownload for Mac

Linux (Beta)

Download for Ubuntu 24.04 (Beta x64)Download for Steam Machine (SteamOS) (Beta x64)

Create a 3D Gaussian scene from your camera recording

Turn a synced Arcturus Vision Camera MP4 into a 3D Gaussian scene for viewing on Steam Frame in Arcturus Vision Gaussian XR.

Spirula Studio: GUI-based

Download for WindowsDownload for macOS

Windows 10/11 · 64-bit · Vulkan-capable GPU with current drivers · Unsigned ZIP

Mac: Apple Silicon · macOS 15+

On Windows, unzip the download and open spirula.exe. On Mac, open the DMG and drag Spirula Studio into Applications.

  1. Copy the original camera MP4 to your computer with Arcturus Vision Sync.
  2. Drag and drop the MP4 into Spirula. Choose a new dataset folder and built-in reconstruction, then prepare the dataset.
  3. Start training. When it finishes, copy splat.ply and splat.xrSceneTransform.json from the final checkpoint folder to /home/steamos/Documents/GaussianXR/ on Steam Frame, then open Gaussian XR to view your scene. To keep several scenes, rename both files to the same new base name (for example, kitchen.ply and kitchen.xrSceneTransform.json).

Linux

The scripts extract and tone-map synchronized stereo images, reconstruct a scene aligned to the recording’s calibration and trajectory, and train at half resolution with a limit of 1.5 million Gaussians. On Linux, Spirula uses its native reconstruction; the LichtFeld workflow uses COLMAP.

  1. 1. Download and unpack the scripts

    Download the AV1 3DGS Pipeline

    Unzip the download and open a terminal in the folder containing av1_to_3dgs.py. Then follow the instructions for your platform below.

  2. 2. Set up and run your workflow

    Spirula Studio

    LinuxNVIDIA / AMD
    Requirements

    NVIDIA or AMD GPU with a working Vulkan driver.

    Build on Linux

    Build Spirula from source using the Linux setup in the downloaded scripts README, then install uv. The pinned Linux build has not yet been revalidated; a packaged Linux installer is coming soon.

    Spirula Studio on GitHubuv
    Optional: export to SOG

    For SOG export, install Node.js and then SplatTransform:

    Node.js
    npm install -g @playcanvas/splat-transform

    Run

    Choose your synced camera MP4 and a new output folder. From the unpacked scripts folder, replace recording.mp4 and output-directory with your paths, then run:

    Train and export PLY (no Node.js required)
    uv run av1_to_3dgs.py recording.mp4 output-directory \
    --backend spirula --sfm-backend spirula --no-undistort \
    --spirula /path/to/spirula-studio/build_vulkan/spirula \
    --iterations 30000

    Train and export PLY + SOG
    uv run av1_to_3dgs.py recording.mp4 output-directory \
    --backend spirula --sfm-backend spirula --no-undistort \
    --spirula /path/to/spirula-studio/build_vulkan/spirula \
    --iterations 30000 --splat-transform splat-transform

    Point --spirula to your source-built executable.

    LichtFeld Studio

    LinuxNVIDIA
    Requirements

    For the LichtFeld Studio workflow, you will need:

    • An NVIDIA GPU for LichtFeld Studio training: compute capability 7.5 or newer (GTX 16-series, RTX 20-series and up), with NVIDIA driver 570 or newer. LichtFeld Studio requirements
    • uv
    • COLMAP — Available on your PATH as colmap
    • LichtFeld Studio — Available on your PATH as LichtFeld-Studio
    Run

    Choose a synced Arcturus Vision Camera MP4 and a new output folder. Replace recording.mp4 and output-directory with your paths, then run:

    uv run av1_to_3dgs.py recording.mp4 output-directory

    LichtFeld Studio opens so you can inspect training. Add --headless to run without its interface.

  3. 3. Output

    Without SOG export, Spirula saves output-directory/spirula/scene.ply. With SOG export, it also creates output-directory/scene.sog.

    With LichtFeld or optional Spirula SOG export, your model is saved to output-directory/scene.sog.

LLM prompt

Paste this into an AI assistant with terminal access to work through the full workflow.

View prompt

Turn my Arcturus Vision Camera recording into a 3D Gaussian model for Steam Frame using the instructions below. Ask for my synced MP4, a new output folder, and my computer platform if missing. On Windows or an Apple Silicon Mac with macOS 15 or later, use the Spirula Studio desktop download and drag-and-drop workflow above; the scripts are documented for Linux only. On Linux, Node.js and SplatTransform are optional and needed only for SOG export; omit --splat-transform for PLY-only output. On Linux with AMD, use the documented Spirula Vulkan source-build route. On Linux with NVIDIA, offer Spirula or the documented LichtFeld Studio baseline. The pinned Linux Spirula build has not yet been revalidated. Download and inspect the scripts ZIP and its README, install the appropriate prerequisites, and run the documented command. Preserve my recording and existing outputs; use a new output directory for changed settings and respect the documented resume limitations. Verify the resulting PLY/SOG and report their paths. Help transfer the SOG to the supported Gaussian XR scene folder on Steam Frame and verify the copied file and rendering. Carry out the steps you can and identify anything I need to do. https://arcturus.vision/apps#create-gaussian-model

Need a hand getting started?

Visit support