Why Nuitka Beats PyInstaller: Faster, Smaller Python Executables

1. PyInstaller vs. Nuitka Experience

The project required turning Python code into an executable, leading to the discovery of two packaging tools: PyInstaller and Nuitka.

Hide source code : PyInstaller encrypts the source with a key, while Nuitka compiles Python to C++ (producing binary .pyd files) to prevent decompilation.

Easy portability : Both tools allow the resulting exe to run without installing Python or third‑party packages.

1.2 Personal impressions

PyInstaller experience:

The generated exe for a deep‑learning project was nearly 3 GB because PyInstaller bundles the entire runtime.

Packaging and startup were both extremely slow.

Nuitka experience:

The same project produced an exe of only about 7 MB.

Packaging completed in under a minute and the executable started instantly.

2. Installing and Using Nuitka

2.1 Installation

Install via pip: pip install Nuitka Download a C++ compiler such as Visual Studio 2019 (MSVC) or MinGW‑64.

2.2 Build command

For projects with many third‑party dependencies (e.g., torch, tensorflow, cv2, numpy, pandas), it is best to compile only the project's own code to C++ and leave the large libraries untouched.

Example directory structure (using a PyQt5 UI):

├─utils//source folder1
├─src//source folder2
├─logo.ico//demo icon
└─demo.py//main file

Run the following command to build the executable (debug mode):

nuitka \
  --standalone \
  --show-memory \
  --show-progress \
  --nofollow-imports \
  --plugin-enable=qt-plugins \
  --follow-import-to=utils,src \
  --output-dir=out \
  --windows-icon-from-ico=./logo.ico demo.py

Explanation of the most important options: --standalone: creates a portable package that runs on other machines without Python installed. --show-memory --show-progress: displays compilation progress and memory usage. --nofollow-imports: skips compiling imported libraries such as keras or numpy. --plugin-enable=qt-plugins: enables the Qt plugin needed for PyQt5. --follow-import-to=utils,src: specifies the two folders containing the project's own source code to be compiled. --output-dir=out: sets the output directory. --windows-icon-from-ico=./logo.ico: sets the executable's icon.

After about one minute of compilation, the output directory contains:

├─utils//source folder1
├─src//source folder2
├─out//generated files
│   ├─demo.build
│   └─demo.dist
│       └─demo.exe // final executable
│       ├─logo.ico // icon
│       └─demo.py // original script

If the executable reports missing modules (e.g., no module named torch, cv2, tensorflow), copy the corresponding package folders from the Python site‑packages directory into demo.dist. After that, the exe runs correctly.