Repolode
Submit
harry7557558

harry7557558/spirula-studio

Cross-vendor 3D Gaussian Splatting trainer that turns photos or video into splats and textured meshes, running on Vulkan or CUDA without Python.

View on GitHub
Stars
2,191
Forks
213
Open issues
70
Last commit
yesterday

What is it?

Spirula Studio trains 3D Gaussian Splatting models from raw photos or video, covering structure-from-motion, training, and mesh generation in a single self-contained binary. It does not require Python, PyTorch, or a separate COLMAP installation.

The main Vulkan backend runs on NVIDIA, AMD, Intel, and Apple GPUs across Windows, Linux, and macOS. A CUDA backend is also available for NVIDIA GPUs. The project highlights quantized training that handles up to 10 million SH3 Gaussians in 8 GB of VRAM, and it supports fisheye and 360° cameras, including equirectangular datasets without undistortion.

The built-in pipeline includes SfM, frame extraction from videos, AI masking via SAM checkpoints, depth and normal output, and a web viewer that can be forwarded over SSH for remote GPU training.

Key features

  • 01Runs on NVIDIA, AMD, Intel, and Apple GPUs through Vulkan compute.
  • 02Trains up to 10 million SH3 Gaussians within 8 GB of VRAM using quantized training.
  • 03Supports 360° and equirectangular camera datasets directly.
  • 04Includes built-in SfM, video frame extraction, and AI masking.
  • 05Exports depth, normals, and textured meshes.
  • 06Offers a GUI and a CLI with a browser-based viewer for remote training.

Use cases

  • Reconstructing 3D scenes from phone or consumer video without installing Python or COLMAP.
  • Training splats on AMD, Intel, or Apple Silicon machines where CUDA is unavailable.
  • Working with 360° or fisheye footage without a separate undistortion step.
  • Running training on remote GPU servers and monitoring progress over SSH.

Quick start

  1. 1

    Download the binary for your platform from the Releases page

  2. 2

    Clone the repository and enter its folder (Linux build from source, Vulkan backend)

    cd spirula-studio/
  3. 3

    Run the Linux build script with the Vulkan backend

    bash build_develop.bash -DSS_BACKEND=vulkan -DSS_ENABLE_PATENTED=ON

Things to know

  • Licensed under GPL-3.0, a copyleft license that requires derivative works to be distributed under the same terms.
  • Building from source needs the Vulkan SDK installed first.
  • The project is maintained almost entirely by one person, so responses to issues and PRs can be slow.

In collections

Computer vision libraries & tools

Topics

#3d-gaussian-splatting#3d-reconstruction#computer-vision#vulkan#cuda#structure-from-motion#mesh-generation#vram-optimization#360-camera#3dgs

FAQ

What is spirula-studio?
Cross-vendor 3D Gaussian Splatting trainer that turns photos or video into splats and textured meshes, running on Vulkan or CUDA without Python. Spirula Studio trains 3D Gaussian Splatting models from raw photos or video, covering structure-from-motion, training, and mesh generation in a single self-contained binary. It does not require Python, PyTorch, or a separate COLMAP installation.
Is spirula-studio free to use? What license does it have?
spirula-studio is open source under the GPL-3.0 license. Check the license terms before using it in commercial products.
Is spirula-studio still maintained?
Actively maintained. The last commit to spirula-studio was on Nov 6, 2026.
What can I use spirula-studio for?
Reconstructing 3D scenes from phone or consumer video without installing Python or COLMAP. Training splats on AMD, Intel, or Apple Silicon machines where CUDA is unavailable. Working with 360° or fisheye footage without a separate undistortion step. Running training on remote GPU servers and monitoring progress over SSH.
How do I install spirula-studio?
Download the binary for your platform from the Releases page Clone the repository and enter its folder (Linux build from source, Vulkan backend): cd spirula-studio/ Run the Linux build script with the Vulkan backend: bash build_develop.bash -DSS_BACKEND=vulkan -DSS_ENABLE_PATENTED=ON
Where can I get spirula-studio?
The source code and installation instructions are on GitHub: https://github.com/harry7557558/spirula-studio

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects