gltf-draco-transcoder 0.3.0

Python package for GLTF Draco transcoding with C++ backend

gltf_draco_transcoder

A high-performance Python package for compressing and decompressing glb files using the Draco 3D data compression library. This package provides both a Python API and a command-line interface for efficient 3D model compression.

Features

• Compression: Compress glb files using Draco compression to reduce file sizes significantly
• Decompression: Decompress Draco-compressed glb files back to standard glTF format
• High Performance: C++ backend with Python bindings for maximum performance
• Cross-platform: Supports Windows, macOS (ARM64), and Linux (x86_64 and ARM64)
• Configurable: Fine-tune compression quality with quantization and compression level settings
• Easy Integration: Simple Python API and command-line interface

Live Demo

Try the online demo to see Draco compression in action on glTF files. The demo uses the WebAssembly (WASM) port of the same C++ codebase as this Python package.

Installation

Install from PyPI:

pip install gltf_draco_transcoder

Usage

Python API

Basic Compression

from gltf_draco_transcoder import compress_gltf, decompress_gltf

# Compress a glb file
compressed_data = compress_gltf('input.glb')

# Save compressed data
with open('output_compressed.glb', 'wb') as f:
    f.write(compressed_data.getvalue())

# Decompress back to glb
decompressed_data = decompress_gltf('output_compressed.glb')
with open('output_decompressed.glb', 'wb') as f:
    f.write(decompressed_data.getvalue())

Advanced Compression with Custom Options

from gltf_draco_transcoder import compress_gltf

# Compress with custom quantization and compression level settings
# qp: quantization_position, qn: quantization_normal, cl: compression_level
compressed_data = compress_gltf('input.glb', qp=14, qn=10, cl=9)

Command Line Interface

Compress a glb file:

python -m gltf_draco_transcoder input.glb output_compressed.glb

This will display compression statistics including file sizes and compression ratio.

Compression Options

The compress_gltf function accepts the following parameters:

• Quantization Bits (default values shown):

  • qp=11: Position quantization
  • qt=10: Texture coordinate quantization
  • qn=8: Normal vector quantization
  • qc=8: Color quantization
  • qtg=8: Tangent quantization
  • qw=8: Weight quantization
  • qg=8: Generic attribute quantization

• Compression Level:

  • cl=7: Compression effort (0-10, higher = better compression but slower)

Higher quantization bits = better quality but larger file sizes. Higher compression levels = better compression ratios but slower processing.

Development

Building from Source

This project uses scikit-build-core with CMake and vcpkg for dependency management.

Prerequisites

• Python 3.10+
• CMake 3.15+
• Git
• vcpkg (automatically installed on Linux via CI scripts)

Build Steps

# Clone the repository
git clone https://github.com/gongfan99/gltf_draco_transcoder.git
cd gltf_draco_transcoder

# Install build dependencies
pip install scikit-build-core setuptools-scm

# Build the package
python -m build

Project Structure

src/gltf_draco_transcoder/
├── __init__.py          # Main API exports
├── loader.py            # Python wrapper and ctypes bindings
├── py.typed             # Type hints marker
└── _version.py          # Version information (auto-generated)

cpp/
├── draco_transcoder_c_api.*    # C API wrapper
└── draco_transcoder_lib.*      # Core C++ implementation

vcpkg-overlays/draco/           # Custom Draco build configuration
vcpkg-triplets/                 # Platform-specific build triplets

Platform Support

• Linux: x86_64 and ARM64 (manylinux/musllinux wheels)
• macOS: ARM64
• Windows: x86_64 (win_amd64 wheels)

Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues on GitHub.

License

This project is licensed under the Apache License Version 2.0 - see the LICENSE file for details.

Credits

• Built on the Draco 3D data compression library
• Uses tinygltf for glTF file I/O
• Uses vcpkg for dependency management
• Packaging with scikit-build-core and cibuildwheel