ntsc-encoder (pybind11)

This directory builds the ntsc_encoder Python extension module using C++/pybind11.

Requirements

• Python 3.9+
• A C++ compiler + CMake 3.18+
• No OpenCV dependency is required to build/install this wheel. The input is a NumPy HxWx3 uint8 array (e.g. produced by cv2.imread() + cv2.cvtColor).
• Build tools:
  • python -m pip install -U build twine

Build a wheel/sdist locally

From this directory (Code/Encoder):

python -m pip install -U pip
python -m pip install -U build twine scikit-build-core pybind11 numpy
python -m build

Artifacts will appear in dist/.

Check metadata:

twine check dist/*

Upload to PyPI

Recommended: upload to TestPyPI first.

# TestPyPI
python -m twine upload -r testpypi dist/*

# PyPI
python -m twine upload dist/*

You’ll need an API token configured in ~/.pypirc or exported via environment variables.

Building wheels for multiple platforms (recommended)

To avoid requiring end users to have a compiler + OpenCV dev libs, publish prebuilt wheels. Typical approach: GitHub Actions + cibuildwheel.

High-level steps:

1. Add cibuildwheel workflow
2. Build wheels for Linux (manylinux), macOS, Windows
3. Upload wheels + sdist with twine

If you want, I can add a ready-to-use GitHub Actions workflow for cibuildwheel.

Troubleshooting

If pip install ntsc-encoder succeeds but import ntsc_encoder fails, it usually means the wheel did not contain the compiled extension (or it was installed into the wrong location). For scikit-build-core, ensure CMakeLists.txt has install(TARGETS ntsc_encoder ... DESTINATION ${SKBUILD_PLATLIB_DIR}).