19 releases
| 0.4.3 | Mar 15, 2024 |
|---|---|
| 0.4.2 | Mar 15, 2024 |
| 0.3.6 | Sep 6, 2023 |
| 0.3.1 | May 8, 2023 |
| 0.1.7 |
|
#104 in Audio
928 downloads per month
325KB
6.5K
SLoC
pointillism
A compositional library for musical composition.
Examples
If you want to see pointillism in action and what it's capable of, run the examples in the
examples folder. There's also many simple examples scattered throughout the source code, showing
off different features.
For a starting example, see the create docs.
Note: Some examples may be loud, dissonant, and/or jarring. Hearing discretion is advised.
Design
The default way in which pointillism outputs audio is by writing sample by sample into a 32-bit
floating point .wav file. Internal calculations use 64-bit floating points.
For convenience, the Signal trait is provided. Types implementing this trait generate sample
data frame by frame. If the type also implements SignalMut, it can be advanced or retriggered.
Signals may be composed to create more complex signals, using for instance the eff::MapSgn and
eff::MutSgn structs. Moreover, you can implement the Signal and SignalMut traits for
your own structs, giving you vast control over the samples you're producing.
Naming scheme
The pointillism code has a lot of moving parts, and a bunch of similarly named types. Because of
this, we rely on the prelude to categorize things neatly.
Every type has a three-letter namespace which helps categorizes it. The main namespaces are as follows:
| Namespace | Full Name | Contents |
|---|---|---|
[buf] |
buffers |
Audio buffers and associated traits. |
[crv] |
curves |
Basic oscillator shapes, and builder methods for more complex ones (in the future). |
[ctr] |
control |
Control structures, which allow for events to happen at specified time intervals. |
[eff] |
effects |
For effects, meaning types that alter other signals. |
[map] |
map |
Basic maps and associated traits. |
[gen] |
generators |
Types that generate a signal "on their own". This includes the basic oscillators like gen::Loop and gen::Once. |
[sgn] |
signal |
Traits on signals, including the basic Signal and SignalMut. |
[smp] |
smp |
Basic traits and types for sample types, including smp::Mono and smp::Stereo. |
[rtn] |
routing |
Structures for mixing or combining different signals together. |
[unt] |
units |
Different units for musical measurement, and associated arithmetical boilerplate. |
Note that traits are always imported when the prelude is imported. This simplifies some complex
impl declarations, and also makes the trait methods available whenever.
Some of these namespaces also contain further nested namespaces, almost always three letters. See the documentation for the full breakdown.
Compile-time
You can think of pointillism as a compile-time modular synthesizer, where every new struct is its own module.
Advantages of this design are extensibility and generality. It's relatively easy to create a highly customizable and complex signal with many layers by composing some functions together.
The downside is that these synths end up having unwieldy type signatures. Moreso, it's really hard to build synths in real time.
Features
The project uses the following features:
| Feature | Enables |
|---|---|
hound* |
Saving songs as WAV files. |
cpal |
Playing songs in a dedicated thread. |
midly |
Reading and playing back MIDI files. |
human-duration* |
Pretty-printing for the unt::RawTime type. |
* Features marked with an asterisk are enabled by default.
Goals
Future goals of pointillism are:
- (Better) algorithmic reverbs
- Limiters, compressors, sidechaining
- Me making a whole album with it :D
Disclaimer
This is a passion project made by one college student learning about DSP. I make no guarantees on it being well-designed, well-maintained, or usable for your own goals.
If you just want to make music with code, and especially if you enjoy live feedback, SuperCollider and Pure Data will most likely be better alternatives for you.
That said, if you happen to stumble across this and make something cool, please let me know!
License: MIT OR Apache-2.0
Dependencies
~0.7–29MB
~438K SLoC