In this post, I'll share the Python toolkit I use to create various tools for game audio implementation.
DISCLAIMER: I'm not a professional programmer in any possible way. All these things I'll be talking about are not Python's best practices or universally-good pieces of advice. These are simply the tools that I, being a Technical Audio Designer, personally found useful for my job.
So, Python for Game Audio?
I believe that Python is one of the best programming languages a Technical Audio Designer can learn. Not only because it is easy to learn, has a big community, is well-supported, and has a lot of libraries, but also because it easily interfaces with the industry-standard tools, such as:
- Wwise (via WAAPI Client for Python),
- Unreal Engine 4 (via Unreal Python API),
- REAPER (via ReaScript Python API or Reapy library),
- Google Sheets (via Google Sheets API),
- and many more.
I'm using Python to automate a broad range of tasks, and I can't imagine doing my job without it.
Here I'll be talking about the tools I use to write Python code.
Well, to program with Python, you'll need Python.
There are two main versions of Python: Python 2 and Python 3. I strongly recommend using Python 3. Here's why.
If you're on Windows, I'll suggest getting Python 3 from its official website. If you're on macOS, it's better to use a package manager such as Homebrew since it'll allow you to manage multiple versions of Python easily. On Windows, you may also use Chocolatey package manager, though I haven't tried it for Python yet.
PyCharm Community by JetBrains is an IDE (integrated development environment, which is a fancy way of saying "text editor that can do a lot of stuff"). It's free; it's fast; it's awesome.
There are two versions of PyCharm: Community and Professional. I use the Community and don't ever feel the need to go with the Professional.
There are some folks (usually the ones with extensive programming background) who prefer "simpler" (not really, in my opinion) and more configurable editors such as Atom, Sublime Text, Visual Studio Code, or even Vim. However, I believe that PyCharm is simply the best tool to do the job quickly and comfortably, without digging into all the settings too much or learning many weird keyboard shortcuts.
TabNine is an autocomplete plugin for PyCharm. It uses Ai and deep learning to predict what I would write next and suggest that to me automagically.
While there is a perfectly good autocomplete already built into PyCharm, TabNine takes it to another level by providing the autofill for the whole lines of code, strings, and other code parts that I would usually be writing by hand. It's hard to overestimate how much time it saves. Here is an example:
Let me explain what's going on in this picture. I wrote a single line,
first_thing = "String number one". When I started to write
seco, not only TabNine suggested continuing it as a
second_thing, it also guessed that the
second_thing's value would be
"String number two", and I never mentioned the word "two" anywhere in my code. It understands the underlying semantics of my code, which makes it super-awesome. It also learns from my code (doing it privately on my computer), getting better the more I use it.
In this part, I'll be talking about Python libraries I use and love the most. Most of them don't work directly with audio, but I still find them incredibly helpful. I'll keep the library's descriptions short so that this post won't end up too big.
Pathlib is the only library in this list that's a part of The Python Standard Library (which means you already have it, if you have Python).
It's a handy library for everything related to paths, files, folders, and filesystem. There are some "older" ways of working with paths in Python, such as
os.path, though I find Pathlib much more convenient to use in almost all cases.
Tip: you can concatenate paths with
/ operator (
combined_path = first_path / second_path), which I find pretty cool.
Loguru is an easy-to-use logging library and a better alternative to using
print() to debug the code and show valuable info on the screen. Here is how it looks:
NumPy (Numerical Python) is an enormously powerful library for working with n-dimensional arrays. Why is it relevant? Because when I'm working with audio, the chances are that the audio data I'm working with is represented as an n-dimensional array. While Python supports a lot of kinds of arrays out-of-the-box, NumPy does it much faster, easier, and has a whole lot of additional tools. Also, there are some pretty fast math functions in Numpy that are not only working with Numpy arrays but also with the usual float numbers.
Whenever I need to do some audio analysis or offline DSP processing in Python, I'll probably end up using NumPy or SciPy, either directly or through other higher-level libraries that use them under the hood.
Reapy is a ReaScript Python API wrapper. It makes it possible to write scripts for the REAPER in Python, using my own Python 3 interpreter, not the one that should be run from the REAPER itself. In other words, I can use a lot more different libraries and write REAPER scripts much more comfortably.
Requests is one of the most popular Python libraries ever! And I use it for everything related to HTTP requests and getting something from the internet.
Pandas is a data manipulation library that I use for data that's somewhat similar to our usual Excel spreadsheets. When I have something that looks like a table with rows, columns, and different data types - I usually go with Pandas.
Librosa is an audio analysis library. It's a bit higher-level than SciPy, thus a bit easier to use.
P4Python is a Perforce API library that helps me build workflows that use the Perforce version control system.
Colorama is a library that allows me to use colors in my program's output.
Pyfiglet is an ASCII art library. I use it to draw beautiful welcome messages for my CLI programs. Like this:
Really important advice
Always use type hints in Python.