Gaussian Adaptive Streamer is a prototype system for adaptive streaming of 3D Gaussian Splatting scenes over modern web transport protocols. The project combines HTTP/3-based delivery, DASH-style adaptive streaming, and server-side Gaussian model handling to efficiently stream large neural rendering datasets to a client.Gaussian Adaptive Streamer

Requirements

• Nvidia GPU with driver.
• FFMPEG with h642_nvenc encoder
• Python dependencies are listed in requirements.txt.

Directory structure

Create a models directory in the project root and place all models inside it:

project_root/
├── static
│    └── models/
│       ├── modelID/
│       │   ├── modelName.ply
│       │   └── preview.jpg
│       └── anotherModelID/
│           ├── anotherModelName.ply
│           └── anotherPreview.jpg
└── requirements.txt

These models and previews will be loaded automatically when starting the server.

Installation

Create a clean Python environment and install the project dependencies.

1. Create environment

conda create -n render python=3.12 -y
conda activate render

2. Install dependencies

python -m pip install --upgrade pip
python -m pip install -r requirements.txt

3. Verify PyTorch + CUDA

python -c "import torch; print('Torch:', torch.__version__); print('CUDA available:', torch.cuda.is_available())"

4. Notes

• The requirements.txt installs CUDA 11.8 PyTorch wheels from the official PyTorch wheel index.

• This project was developed and tested with CUDA 11.8. There is no guarantee that other CUDA versions will work.

• GPU execution also depends on a compatible NVIDIA driver. Systems with incompatible drivers may fail to initialize CUDA.

• The environment was tested with Python 3.12. Other Python versions may not have compatible wheels for the specified PyTorch/CUDA combination.

If GPU support fails, verify:

• nvidia-smi works

• the installed driver version supports CUDA 11.8

• the correct PyTorch CUDA wheels were installed.

Running the Server

Start the streaming server:

python http3_server.py --certificate certificates/ssl_cert.pem --private-key certificates/ssl_key.pem

Start Google Chrome with flags:

Linux:

 google-chrome \
  --enable-experimental-web-platform-features \
  --ignore-certificate-errors-spki-list=BSQJ0jkQ7wwhR7KvPZ+DSNk2XTZ/MS6xCbo9qu++VdQ= \
  --origin-to-force-quic-on=localhost:4433 \
  https://localhost:4433/models-ui

If super resolution is not working, try:

google-chrome \
  --enable-experimental-web-platform-features \
  --enable-unsafe-webgpu \
  --ignore-gpu-blocklist \
  --ignore-certificate-errors-spki-list=BSQJ0jkQ7wwhR7KvPZ+DSNk2XTZ/MS6xCbo9qu++VdQ= \
  --origin-to-force-quic-on=localhost:4433 \
  https://localhost:4433/models-ui

Note:

• For trying the experimental version with dash.js as player type instead of /models-ui, go to /player-dash.
• Close Google Chrome before running this command.

Preview

Model Selection

Viewer

Experiment