AirVis
Credits
We gratefully acknowledge the following open-source projects that support the AirVis experience.
Used in AirVis
These open-source projects are directly used in the AirVis workflow and experience.
gsplat by nerfstudio-project
License: Apache License 2.0
A Gaussian Splatting implementation integrated with Nerfstudio.
COLMAP by colmap
License: BSD 3-Clause License
A general-purpose Structure-from-Motion and Multi-View Stereo pipeline.
spark by sparkjsdev
License: MIT License
An advanced 3D Gaussian Splatting renderer for THREE.js.
MediaPipe Selfie Segmentation by Google
License: Apache License 2.0
A real-time person segmentation solution used for foreground and background masking workflows.
BEN2 Base by Prama LLC
License: MIT License
A foreground segmentation model used for Object Isolation in AirVis Windows workflows.
MODNet by ZHKKKe
License: Apache License 2.0
A real-time portrait matting model used for People Masking in AirVis Windows workflows.
TinyUSDZ
License: Apache License 2.0
A lightweight USD and USDZ library. AirVis acknowledges TinyUSDZ for USDZ loading and conversion in the Windows USDZ/GLB viewer.
cgltf
License: MIT License
A single-file glTF 2.0 parser. AirVis acknowledges cgltf for parsing binary glTF/GLB files in the Windows viewer.
Referenced Projects
These projects are referenced during development or research, but are not directly used in AirVis.
SuperSplat by PlayCanvas
License: MIT License
An open-source editor for 3D Gaussian splats. Referenced for SSOG and SOG workflows and collision generation.
brush by Arthur Brussee
License: Apache License 2.0
Brush is a 3D reconstruction engine using Gaussian splatting.
MetalSplatter by scier
License: MIT License
3D Gaussian Splats using Metal on Apple platforms.
SwiftyCrop by benedom
License: MIT License
A Swift-based image cropping library.
spz by nianticlabs
License: MIT License
.spz is a file format for compressed 3D gaussian splats.
Previously Used
These projects were used in earlier versions of AirVis but are not part of the current product.
Depth-Anything-3
License: Apache License 2.0
DA3-BASE predicts spatially consistent geometry from visual inputs, with or without known camera poses. It was used in earlier versions of AirVis.
License Notes
Each project listed here retains its original license and ownership. We do not claim ownership of any of these projects. Please refer to their respective repositories for detailed license information.
Last updated: 08.20.2026