Tiny USDZ library in C++14

TinyUSDZ is secure, portable and dependency-free(depends only on C++ STL. Other 3rd-party libraries included. Yes, you don't need pxrUSD library!) USDZ/USDC/USDA library written in C++14.

Mid-term todo

• Experimental composition support #25
  • subLayers
  • references
  • payload(no delayed load)
  • inherits
  • variantSet
  • Validate composition is correctly operated.
• Tydra: Handy data structure converter for rendering #31
• MaterialX #86
  • USD + MateriralX + PBR rendering example using https://github.com/lighttransport/pbrlab
• Improve support for Blender USD export/import
• tusdview(TinyUSDZ version of usdview)

Build status

	Linux	Windows	macOS	iOS	Android
dev

Supported platforms

Python binding(testing. currently not working)

https://pypi.org/project/tinyusdz/

Python binding is very early alpha testing stage. Not working at the moment.

You can install Python prebuilt wheel using

$ python -m pip install tinyusdz

NOTE: Windows ARM64 binary is provided using cross-compiling. Its not well tested.

Status

TinyUSDZ is in v0.8.0 release candidate. Core loading feature(both USDA and USDC) is now working and production-grade(And no seg fault for corrupted USDA/USDC/USDZ input).

v0.8.0 is Flattened scene only(i.e, USDA/USDC generated by using pxrUSD's usdcat --flatten or USDZ scene). Composition features are work-in-progress(experimental Composition feature support in v0.8.0. Better composition feature in next major release v0.9.0(Q4/2023 expected) planned)

Remaining tasks for v0.8.0 release are writing examples, demos and utility functions(Tydra. Especially access to Material/Shader attributes).

• USDZ/USDC(Crate) parser
  • USDC Crate version v0.8.0(most commonly used version as of 2022 Nov) or higher is supported.
• USDZ/USDC(Crate) writer (Work-in-progress)
• USDA parser(Hand-written from a scratch. No Bison/Flex dependency!)
• USDA writer
• Support basic Primitives(Xform, Mesh, BasisCurves, etc.), basic Lights and Shaders(UsdPreviewSurface, UsdUVTexture, UsdPrimvarReader)
• Experimental support of some Composition features #25
  • subLayers
  • references
  • payload
  • inherits
  • variants
  • specializes

Please see doc/status.md for more details

• Basic C API(c-tinyusd) for language bindings
  • examples/c_api_example
  • Basic Python binding
• Write simple SDL viewer example(2023 Summer expected)
• Write iOS and Android example(2023 Summer expected)
• Vulkan or OptiX/HIP RT raytracing viewer example
• USD <-> glTF converter example
  • There is an independent work of USD to glTF binary GLB converter: https://github.com/fynv/usd2glb
• Web demo with Three.js?
  • Three.js started to support USDZ with Ascii format, but no USDC support yet: mrdoob/three.js#14219

Discussions

We've opened Github Discussions page! https://github.com/syoyo/tinyusdz/discussions

Security and memory budget

TinyUSDZ has first priority of considering security and stability.

USDZ(USDC) is a binary format. To avoid out-of-bounds access, out-of-memory, and other security issues when loading malcious USDZ(e.g. USDZ file from unknown origin), TinyUSDZ has a memory budget feature to avoid out-of-memory issue.

To limit a memory usage when loading USDZ file, Please set a value max_memory_limit_in_mb in USDLoadOptions.

TinyUSDZ source codes(and some external third party codes) are also checked by Address Sanitizer, CodeQL and Fuzzer.

Fuzzer

See tests/fuzzer . For building fuzzer tests, you'll need Meson and Ninja.

Web platform(WASM) and sandboxed environment(WASI)

If you need to deal with arbitrary USD files from unknown origin(e.g. from internet, NFT storage. Whose may contain malcious data), it is recommended to use TinyUSDZ in sandboxed environment(RunC, FlatPak, WASI(WASM)). Run in WASI is recommended at the moment.

TinyUSDZ does not use C++ exceptions and can be built without threads. TinyUSDZ supports WASM and WASI build. So TinyUSDZ should runs well on various Web platform(WebAssembly. No SharedArrayBuffer, Atomics and WebAssembly SIMD(which is not yet available on iOS Safari) required) and sandboxed environment(WASI. Users who need to read various USD file which possibly could contain malcious data from Internet, IPFS or blockchain storage).

See sandbox/wasi/ for Building TinyUSDZ with WASI toolchain.

Tydra

USD itself is a generic container of 3D scene data.

Tydra is an interface to Renderers/Viewers and other DCCs. Tydra may be something like Tiny version of pxrUSD Hydra, but its API is completely different. See src/tydra/README.md for the background.

• Image color space
  • sRGB
  • Linear
  • Rec.709
  • Partial support of OCIO(OpenColor IO) through TinyColorIO https://github.com/syoyo/tinycolorio . Currently SPI3DLut only.
• More details are T.B.W.

Notice

TinyUSDZ does not support Reality Composer file format(.reality) since it uses proprietary file format and we cannot understand it(so no conversion support from/to Reality also).

Commercial support

TinyUSDZ focuses on loading/writing USDA/USDC/USDZ functionalities. Example viewer is just for demo purpose. syoyo does not provide commercial support as an individual.

If you need commercial support, eco-system development(e.g. plug-ins, DCC tools on top of TinyUSDZ) or production-grade USDZ model viewer(e.g. embed TinyUSDZ to your AR app, 3D NFT Android mobile viewer capable of displaying (encrypted) USDZ model), please contact Light Transport Entertainment, Inc. : https://goo.gl/forms/1p6uGcOKWGpXPHkA2

We have a plan to manage TinyUSDZ project under Light Transport Entertainment Inc. (By relicensing to Apatch 2.0)

Projects using TinyUSDZ

• usd2glb: USD to glTF 2.0 GLB converter https://github.com/fynv/usd2glb

Other related projects

• UsdzSharpie: C# Simple implementation of Usdz file format ( https://github.com/UkooLabs/UsdzSharpie )
• TinyGLTF: glTF 2.0 loader/saver ( https://github.com/syoyo/tinygltf )
• USD-build-aarch64: Full USD build for AARCH64(Linux and Android): https://github.com/syoyo/USD-build-aarch64
• BlenderUSDZ: It contains their own Python USDC parser/serializer. https://github.com/robmcrosby/BlenderUSDZ

Supported platforms

• Linux 64bit or later
  • ARM AARCH64
  • x86-64
  • RISC-V(Should work)
  • SPARC, POWER(Big endian machine). May work(theoretically)
• Android arm64v8a
• iOS
• macOS(Arm, x86-64)
• Windows 10 64bit or later
  • Windows ARM
  • clang-cl + MSVC SDK cross compile
• WebAssembly
  • Emscripten
    • See examples/sdlviewer/ example.
• WASI(through WASI toolchain)
  • See sandbox/wasi

Requirements

• C++14 compiler
  • gcc 4.9 or later
  • Visual Studio 2019 or later(2017 may compiles)
    • VS2019 16.10 or later you can use CMakePresets.json for easier building.
    • Can be compiled with standalone MSVC compilers(Build Tools for Visual Studio 2019)
  • clang 3.4 or later https://clang.llvm.org/cxx_status.html
  • llvm-mingw(clang) supported
  • MinGW gcc supported, but not recommended(You may got compilation failure depending on your build configuration: #33 , and linking takes too much time if you use default bfd linker.). If you want to compile TinyUSDZ in MinGW environment, llvm-mingw(clang) is recommended to use.

Compilation with C++17 is also supported.

Build

Integrate to your app

If you are using CMake, just include tinyusdz repo with add_subdirectory and set include path to <tinyusdz>/src We recommend to use CMake 3.24 or later. (Mininum requirement is 3.16)

...

# TinyUSDZ examples, tests and tools builds are disabled by default when
# tinyusdz is being built as a library with `add_subdirectory`
add_subdirectory(/path/to/tinyusdz tinyusdz)

target_include_directories(YOUR_APP PRIVATE "/path/to/tinyusdz/src")

# Namespaced static library target `tinyusdz::tinyusdz_static` is provided.
# At the moment we recommend to use static build of TinyUSDZ. 
# You can use alias target `tinyusdz_static` also for legacy cmake version. 
target_link_libraries(YOUR_APP PRIVATE tinyusdz::tinyusdz_static)

# For TinyUSDZ DLL(shared) library target, you can use
# `tinyusdz` library target  

Another way is simply copy src folder to your app, and add *.cc files to your app's build system. All include paths are set relative from src folder, so you can just add include directory to src folder.

See <tinyusdz>/CMakeLists.txt and examples/sdlviewer/CMakeLists.txt for details.

TinyUSDZ does not generate any header files and source files before the build and after the build(before the installation stage), so you don't need to take care of any pre-processing and post-processing of source tree. For example, USD Ascii parser uses hand-written C++ code so no Bison/flex/PEG processing involved.

It may not be recommend to use tinyusdz as a git submodule, since the repo contains lots of codes required to build TinyUSDZ examples but these are not required for your app.

Compiler defines

Please see CMake build options and CMakeLists.txt. In most case same identifier is defined from cmake build options: For example if you specify -DTINYUSDZ_PRODUCTION_BUILD=1 for cmake argument, TINYUSDZ_PRODUCTION_BUILD is defined.

CMake

Cmake build is provided.

Linux and macOS

$ mkdir build
$ cd build
$ cmake ..
$ make

Please take a look at scripts/bootstrap-cmake-*.sh for some build configuraions.

Visual Studio

Visual Studio 2019 and 2022 are supported.

CMakePresets.json is provided for easier build on Visual Studio 2019 and Visual Studio 2022, but has lot of limitations(and seems parallel build is not working well so build is slow).

If you want flexibility, ordinary cmake .sln generation approach by invoking vcsetup.bat recommended. (Edit VS version in vcsetup.bat as you with before the run)

LLVM-MinGW build

MinGW native and cross-compiling example using llvm-mingw(clang) is provided. See scripts/bootstrap-cmake-mingw-win.sh and scripts/bootstrap-cmake-llvm-mingw-cross.sh for details.

One of benefit to use llvm-mingw is address sanitizer support on Windows app.

To run app(.exe, you'll need libunwind.dll and libc++.dll on your working directory or search path)

For Windows native build, we assume ninja.exe is installed on your system(You can use it from Meson package)

CMake build options

• TINYUSDZ_PRODUCTION_BUILD : Production build. Do not output debugging logs.
• TINYUSDZ_BUILD_TESTS : Build tests
• TINYUSDZ_BUILD_EXAMPLES : Build examples(note that not all examples in examples folder are built)
• TINYUSDZ_WITH_OPENSUBDIV : Use OpenSubviv to tessellate subdivision surface.
  • OpenSubdiv code is included in TinyUSDZ repo. If you want to use external OpenSubdiv repo, specity the path to OpenSubdiv using osd_DIR cmake environment variable.
• TINYUSDZ_WITH_AUDIO : Support loading audio(mp3 and wav).
• TINYUSDZ_WITH_EXR : Support loading EXR format HDR texture through TinyEXR.
• TINYUSDZ_WITH_PXR_COMPAT_API : Build with pxrUSD compatible API.

clang-cl on Windows

Assume ninja.exe is installed and path to ninja.exe is added to your %PATH%

Edit path to MSVC SDK and Windows SDK in bootstrap-clang-cl-win64.bat, then

> bootstrap-clang-cl-win64.bat
> ninja.exe

Tools and Examples

• tusdcat Parse USDZ/USDA/USDC and print it as Ascii(similar to usdcat in pxrUSD).
• usda_parser Parse USDA and print it as Ascii.
• usdc_parser Parse USDC and print it as Ascii.
• Simple SDL viewer
  • Separated CMake build provided: See Readme
• api_tutorial Tutorial of TinyUSDZ Core API to construct a USD scene data.
• tydra_api Tutorial of TinyUSDZ Tydra API to access/query/convert a USD scene data.

See examples directory for more examples, but may not actively maintained except for the above examples.

USDZ Data format

See prim_format.md and preview_surface.md

Example

Minimum example to load USDA/USDC/USDZ file.

// TinyUSDZ is not a header-only library, so no TINYUSDZ_IMPLEMENTATIONS
#include "tinyusdz.hh"

// Include pprinter.hh and value-pprint.hh if you want to print TinyUSDZ classes/structs/enums.
// `tinyusdz::to_string()` and `std::operator<<` for TinyUSDZ classes/enums are provided separately for faster compilation
#include <iostream>
#include "pprinter.hh"
#include "value-pprint.hh"

int main(int argc, char **argv) {

  std::string filename = "input.usd";

  if (argc > 1) {
    filename = argv[1];
  }

  tinyusdz::Stage stage; // Stage in USD terminology is nearly meant for Scene in generic 3D graphics terminology.
  std::string warn;
  std::string err;

  // Auto detect USDA/USDC/USDZ
  bool ret = tinyusdz::LoadUSDFromFile(filename, &stage, &warn, &err);

  if (warn.size()) {
    std::cout << "WARN : " << warn << "\n";
  }

  if (!ret) {
    if (!err.empty()) {
      std::cerr << "ERR : " << warn << "\n";
    }
    return EXIT_FAILURE;
  }

  // Print Stage(Scene graph)
  std::cout << tinyusdz::to_string(stage) << "\n";
  
  // You can also use ExportToString() as done in pxrUSD 
  // std::cout << stage.ExportToString() << "\n";

  // stage.metas() To get Scene metadatum, 
  for (const Prim &root_prim : stage.root_prims()) {
    std::cout << root_prim.absolute_path() << "\n";
    // You can traverse Prim(scene graph object) using Prim::children()
    // See examples/api_tutorial and examples/tydra_api for details.
  }

  return EXIT_SUCCESS;
}

With Core TinyUSDZ API

Please see api_tutorial

With Tydra API

Please see tydra_api

TODO

Higher priority

• Built-in usdObj(wavefront .obj mesh) support.
  • via tinyobjloader.
• Support Crate(binary) version 0.8.0(USD v20.11 default)
• usdSkel utilities
  • Joint hierachy reconstruction and compute skinning matrix(usdSkel)
  • Blend shapes
    • Basic Blendshapes support
    • In-between blend shapes
• Read USD data with bounded memory size. This feature is especially useful for mobile platform(e.g. in terms of security, memory consumption, etc)
• USDC writer
• Support Nested USDZ
• UDIM texture support
• MaterialX support
  • Parse XML file using tinyxml2

Middle priority

• Composition arcs
• Code refactoring, code optimization

Lower priority

• iOS example?
• Support AR related schema(Game-like feature added by Reality Composer?)
• Audio play support
  • Play audio using SoLoud or miniaudio(or Oboe for Android)
  • wav(dr_wav)
  • mp3(dr_mp3)
  • m4a(ALAC?)
• Viewer with Vulkan API.
• Replace OpenSubdiv with our own subdiv library.
• Write parser based on the schema definition.
• Support big endian architecture.

Python binding and prebuit packages

Python binding and prebuilt packages(uploadded on PyPI) are provided.

See python/README.md and doc/python_binding.md for details.

CI build

CI build script is a build script trying to build TinyUSDZ in self-contained manner as much as possible(including custom Python build)

Linux/macOS

T.B.W.

Windows

Build custom Python,

> ci-build-python-lib.bat

then build TinyUSDZ by linking with this local Python build.

> ci-build-vs2022.bat

Cross compile with clang-cl + MSVC SDK on linux and run it on WINE(No Windows required at all solution!)

clang-cl(MSVC cl.exe) + MSVC SDK cross compile is also supported.

Please take a look at doc/wine_cl.md

You can build pure Windows build of TinyUSDZ on Linux CI machine.

License

TinyUSDZ is licensed under MIT license and Apache 2.0 license. (Doing relicensing from MIT to Apache 2.0. Will be fully relicensed to Apache 2.0 at some point)

Third party licenses

• pxrUSD : Apache 2.0 license. https://github.com/PixarAnimationStudios/USD
• OpenSubdiv : Apache 2.0 license. https://github.com/PixarAnimationStudios/OpenSubdiv
• lz4 : BSD-2 license. http://www.lz4.org
• cnpy(uncompressed ZIP decode/encode code) : MIT license https://github.com/rogersce/cnpy
• tinyexr: BSD license.
• tinyobjloader: MIT license.
• tinygltf: MIT license.
• tinycolorio: MIT license. https://github.com/syoyo/tinycolorio
• stb_image, stb_image_resize, stb_image_write, stb_truetype: public domain.
• dr_libs: public domain. https://github.com/mackron/dr_libs
• miniaudio: public domain or MIT no attribution. https://github.com/dr-soft/miniaudio
• SDL2 : zlib license. https://www.libsdl.org/index.php
• optional-lite: BSL 1.0 license. https://github.com/martinmoene/optional-lite
• expected-lite: BSL 1.0 license. https://github.com/martinmoene/expected-lite
• mapbox/earcut.hpp: ISC license. https://github.com/mapbox/earcut.hpp
• par_shapes.h generate parametric surfaces and other simple shapes: MIT license https://github.com/prideout/par
• staticstruct: MIT license. https://github.com/syoyo/staticstruct
• MaterialX: Apache 2.0 license. https://github.com/AcademySoftwareFoundation/MaterialX
• tinyxml2: zlib license. https://github.com/leethomason/tinyxml2
• string_id: zlib license. https://github.com/foonathan/string_id
• cityhash: MIT license. https://github.com/google/cityhash
• fast_float: Apache 2.0/MIT dual license. https://github.com/fastfloat/fast_float
• jsteeman/atoi: Apache 2.0 license. https://github.com/jsteemann/atoi
• formatxx: unlicense. https://github.com/seanmiddleditch/formatxx
• ubench.h: Unlicense. https://github.com/sheredom/ubench.h
• thelink2012/any : BSL-1.0 license. https://github.com/thelink2012/any
• simple_match : BSL-1.0 license. https://github.com/jbandela/simple_match
• nanobind : BSD-3 license. https://github.com/wjakob/nanobind
• pybind11 : BSD-3 license. https://github.com/pybind/pybind11
• pystring : BSD-3 license. https://github.com/imageworks/pystring
• gulrak/filesytem : MIT license. https://github.com/gulrak/filesystem
• p-ranav/glob : MIT license. https://github.com/p-ranav/glob
• linalg.h : Unlicense. https://github.com/sgorsten/linalg
• mapbox/eternal: ISC License. https://github.com/mapbox/eternal
• bvh: MIT license. https://github.com/madmann91/bvh
• dtoa_milo.h: MIT License. https://github.com/miloyip/dtoa-benchmark
• jeaiii/itoa: MIT License. https://github.com/jeaiii/itoa
• alac: Apache 2.0 License. https://macosforge.github.io/alac/
• UTF8-CPP: BSL 1.0 License. https://github.com/nemtrif/utfcpp
• OpenFBX: MIT License. https://github.com/nem0/OpenFBX
• floaxie: Apache 2.0 License. https://github.com/aclex/floaxie
• boost math sin_pi/cos_pi: BSL 1.0 License. https://www.boost.org/
• Vulkan: MIT License. https://github.com/SaschaWillems/Vulkan
• Metal.cpp: Apache 2.0 License. https://github.com/bkaradzic/metal-cpp https://developer.apple.com/metal/cpp/
• sRGB transform: MIT license. https://www.nayuki.io/page/srgb-transform-library
• virtualGizmo3D: BSD-2 license. https://github.com/BrutPitt/virtualGizmo3D
• nanozlib: Apache 2.0 license. https://github.com/lighttransport/nanozlib
• lz4.py: MIT license. https://github.com/SE2Dev/PyCoD/blob/master/_lz4.py