This is a collection of my "just for curiosity's sake" audio experiments.
This is stuff I work on very sporadically and just for fun. As such, it's not "complete", and never will be - that's not the point. I don't update it regularly, only when I feel like working on it (especially these days - most of my for-fun audio coding right now is in C++).
Not everything fully works in all edge cases; test coverage is spotty, and some of existing the tests fail; and there are quite a few TODOs and FIXMEs. Code quality & style varies, since this started out as a random collection of old files over the previous 5+ years that I figured I should consolidate, much of it originally Python 2. But I figured it's better to put this out there than just letting it languish on my hard drive forever.
For code that's actually meant to be usable, well tested, etc, I'm working on a C++ audio library, some of which is fairly directly ported from this. As of writing this, I haven't made any of that code public yet, but I hope to soon (for some value of "soon" 😉).
Various modules here have main() functions.
For the sake of dealing with relative imports, all modules get run from main.py, with the module name as subcommand -
this will call the module's main().
Many modules also have separate test() and/or plot() functions. You can run these by adding --test or --plot.
Some modules just have main call plot and/or test, but others have unique main functionality.
Run with --help for a list of arguments that main.py takes
Run Dattorro/Griesinger "figure 8" reverb algorithm and analyze the impulse response for density, RT60 etc:
main.py delay_reverb.dattorro_reverb
Process & plot frequency response of linear "4 cascaded poles with feedback" lowpass filter architecture:
main.py filters.cascade
Analyze various soft & hard clipping functions for harmonic distortion:
main.py overdrive.overdrive
Iteratively calculate tanh function with feedback, with both finite and infinite open-loop gain:
main.py overdrive.tanh_fb --plot
Run & analyze various simple compressor algorithm curves:
main.py compression.basic_compressors
As the goal of this code is experimentation, most of it has been written with ease of understanding what's actually going on over performance.
Many of the classes here are written with a "real-time style" API, i.e. you can feed them audio a bit at a time instead of needing to process the whole thing at once. However, the idea behind this was more for the sake of making this easy to port to C/C++ than for actual real-time use.
Yeah, I know. I don't really plan on fixing them - again, the point of this is experimentation, not necessarily code that works in 100% of corner cases.
Yeah, if I was going to start over from scratch, then I'd use pytest. But I didn't when I started a few years ago, and now I have a basic test framework that works and does the (pretty simple) things I need it to.
Yeah, some of this would probably integrate pretty nicely with Spotify's Pedalboard library. But most of this code predates Pedalboard, and I'm not working on this enough anymore that it's really worth adding Pedalboard integration.
Where would the fun be in that?