open-watercolor-sim 0.1.1

A high-performance GPU-accelerated watercolor simulation engine using Taichi.

Open-Watercolor-Sim

Open-Watercolor-Sim is a real-time GPU-accelerated watercolor simulation framework in Python. It provides a physically inspired simulation engine designed for both artistic creation and as a research toolbox for computer graphics and fluid dynamics exploration.

Demo

Features

• GPU Acceleration: Built on Taichi Lang for high-performance physics-based simulation.
• Interactive Controls: Real-time brush interaction with adjustable radius, pressure (pigment load), and water release.
• Physically Inspired Effects: Models complex watercolor behaviors including diffusion, evaporation, edge darkening (coffee-ring effect), and granulation.
• Cross-Platform: Support for multiple GPU backends (CUDA, Metal, Vulkan) enabling real-time performance on Windows, macOS, and Linux.
• Research Toolbox: Easily extensible architecture for testing new fluid advection schemes or pigment interaction models.

Installation

Open-Watercolor-Sim requires Python 3.10 or newer.

pip install open-watercolor-sim

Quickstart

You can launch the interactive viewer using the console entry point:

watercolor-sim

Alternatively, run the module directly:

python -m open_watercolor_sim.viewer

Project Structure

demo/                      # Previews and example outputs
src/open_watercolor_sim/   # Main package
├── viewer.py              # Interactive GGUI-based viewer
└── brush/
    ├── configs.py         # Simulation and artistic parameters
    └── watercolor_engine.py # Core Taichi-based simulation logic

Citation

If you use this framework in your research, please cite it using the following BibTeX:

@software{open_watercolor_sim,
  title  = {Open-Watercolor-Sim: A Real-Time GPU-Accelerated Watercolor Simulation Framework},
  author = {Chen, Shuoqi},
  year   = {2026},
  url    = {https://github.com/shuoqichen/Open-Watercolor-Sim}
}

Note: A DOI-backed citation for the paper-aligned release will be added in a future update.

License

This project is licensed under the Apache License 2.0.