Install

Install uv with curl -LsSf https://astral.sh/uv/install.sh | sh and a matching CUDA Toolkit: 11.8, 12.6, 12.8, or 13.0.

git clone https://github.com/Florian-Barthel/splatviz.git
cd splatviz

Sync the environment with the CUDA extra for your system:

uv sync --extra cu118
uv sync --extra cu126
uv sync --extra cu128
uv sync --extra cu130

Then install the rasterizer:

uv pip install --no-build-isolation git+https://github.com/ashawkey/diff-gaussian-rasterization.git

On Windows, you may also need the 2022 Visual Studio C++ build tools.

winget install Microsoft.VisualStudio.2022.BuildTools

Launch

python run_main.py

If OpenGL picks the wrong version, set:

export MESA_GL_VERSION_OVERRIDE=3.3

Attach to a running 3DGS training process:

python run_main.py --mode=attach

By default, splatviz listens on 127.0.0.1:6007. Override with --host and --port. For live editing during training, see splatviz_network.

Widgets

Edit Run Python just before rendering to modify the Gaussian object live.

Evaluate Inspect render variables and visualize values with histograms.

Recent Features

Version 1.3.0

• 2026-05: Video widget
• 2026-05: Easier install with uv
• 2026-05: External edit text support
• 2026-05: Resizable render window

Version 1.2.0

• 2025-09: Fixed eval widget and rasterizer import
• 2025-09: Simpler camera and file-browser scene loading
• 2025-09: Moved GAN mode to ganviz
• 2025-05: Added GAN mode

Version 1.1.0

• 2024-08: Training widget and live 3DGS attach mode
• 2024-08: Rendering refactor and improved mouse controls
• 2024-07: Slider presets, performance widget, and editor tooltips
• 2024-07: Multi-scene comparison

Citation

If splatviz helps your work, please cite:

@misc{barthel2024gaussian,
    title={Gaussian Splatting Decoder for 3D-aware Generative Adversarial Networks},
    author={Florian Barthel and Arian Beckmann and Wieland Morgenstern and Anna Hilsmann and Peter Eisert},
    year={2024},
    eprint={2404.10625},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}

References

• GUI: pyimgui, imgui_bundle, and ImGUI
• Original visualizer: EG3D
• 3DGS: 3D Gaussian Splatting
• Compression: Self-Organizing Gaussian Grids
• Rasterizer: diff-gaussian-rasterization