Viser is a 3D visualization library for computer vision and robotics in Python.
Features include:
- API for visualizing 3D primitives.
- GUI building blocks: buttons, checkboxes, text inputs, sliders, etc.
- Scene interaction tools (clicks, selection, transform gizmos).
- Programmatic camera control and rendering.
- An entirely web-based client, for easy use over SSH!
The goal is to provide primitives that are (1) easy for simple visualization tasks, but (2) can be composed into more elaborate interfaces. For more about design goals, see the technical report.
Examples and documentation: https://viser.studio
You can install viser
with pip
:
pip install viser # Core dependencies only.
pip install viser[examples] # To include example dependencies.
That's it! To learn more, we recommend looking at the examples in the documentation.
To cite Viser in your work, you can use the BibTeX for our technical report:
@misc{yi2025viser,
title={Viser: Imperative, Web-based 3D Visualization in Python},
author={Brent Yi and Chung Min Kim and Justin Kerr and Gina Wu and Rebecca Feng and Anthony Zhang and Jonas Kulhanek and Hongsuk Choi and Yi Ma and Matthew Tancik and Angjoo Kanazawa},
year={2025},
eprint={2507.22885},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2507.22885},
}
viser
is heavily inspired by packages like
Pangolin,
Dear ImGui,
rviz,
meshcat, and
Gradio.
The web client is implemented using React, with:
- Vite / Rollup for bundling
- three.js via react-three-fiber and drei
- Mantine for UI components
- zustand for state management
- vanilla-extract for stylesheets
Thanks to the authors of these projects for open-sourcing their work!