ideoforms / signalflow Goto Github PK

View Code? Open in Web Editor NEW

167.0 8.0 14.0 9.08 MB

A sound synthesis framework for Python, designed for clear and concise expression of complex musical ideas

Home Page: https://signalflow.dev

License: MIT License

C++ 85.77% C 0.59% Python 12.22% Objective-C++ 0.21% Shell 0.16% CMake 1.06%

audio synthesis sound music composition python cpp dsp sound-synthesis

signalflow's Introduction

SignalFlow

SignalFlow is a sound synthesis framework designed for clear and concise expression of complex musical ideas. It has an extensive Python API, for fluid audio experimentation in iPython/Jupyter, with its core written in portable, hardware-accelerated C++11.

SignalFlow is in beta status. Interfaces may be subject to change.

Example

from signalflow import *

#--------------------------------------------------------------------------------
# An AudioGraph is made up of a network of interconnected Nodes, which generate
# and process audio. 
#--------------------------------------------------------------------------------
graph = AudioGraph()

#--------------------------------------------------------------------------------
# Passing an array of frequencies creates a stereo output.
#--------------------------------------------------------------------------------
sine = SineOscillator([440, 880])

#--------------------------------------------------------------------------------
# Simple attack/sustain/release envelope with linear curves.
#--------------------------------------------------------------------------------
env = ASREnvelope(0.01, 0.1, 0.5)

#--------------------------------------------------------------------------------
# Use standard arithmetic operations to combine signals. When a multi-channel 
# signal is multiplied by a mono signal, the mono signal is auto-upmixed.
#--------------------------------------------------------------------------------
output = sine * env

#--------------------------------------------------------------------------------
# Connect the output to the graph, and begin playback.
#--------------------------------------------------------------------------------
output.play()
graph.wait()

Installation

SignalFlow requires Python 3.8 or above.

In brief, SignalFlow can be installed with pip: pip3 install signalflow.

For complete per-platform instructions and setup guides, see Installing SignalFlow.

Examples

Several example scripts are provided in examples.

To download the examples folder:

import signalflow_examples
signalflow_examples.download_examples()

Documentation

In-depth documentation, including a reference guide to each of the Node classes, can be found at signalflow.dev.

Node classes

The following Node classes are currently included with the base distribution:

Category	Classes
Analysis	CrossCorrelate, NearestNeighbour, OnsetDetector, VampAnalysis
Buffer	BeatCutter, BufferLooper, BufferPlayer, BufferRecorder, FeedbackBufferReader, FeedbackBufferWriter, SegmentPlayer
Buffer: Granulation	SegmentedGranulator, Granulator
Control	MouseX, MouseY, MouseDown
Envelope	ADSREnvelope, ASREnvelope, DetectSilence, Envelope, Line, RectangularEnvelope
FFT	FFTContinuousPhaseVocoder, FFTConvolve, FFTContrast, FFTFlipSpectrum, FFT, FFTNode, FFTOpNode, FFTFindPeaks, IFFT, FFTLPF, FFTNoiseGate, FFTPhaseVocoder, FFTTonality, FFTZeroPhase
Operators	Add, AmplitudeToDecibels, DecibelsToAmplitude, ChannelArray, ChannelCrossfade, ChannelMixer, ChannelSelect, Equal, NotEqual, GreaterThan, GreaterThanOrEqual, LessThan, LessThanOrEqual, Modulo, Abs, If, Divide, FrequencyToMidiNote, MidiNoteToFrequency, Multiply, Pow, RoundToScale, Round, ScaleLinExp, ScaleLinLin, Subtract, Sum, Sin, Cos, Tan, Tanh
Oscillators	Constant, Impulse, LFO, SawLFO, SawOscillator, SineLFO, SineOscillator, SquareLFO, SquareOscillator, TriangleLFO, TriangleOscillator, Wavetable, Wavetable2D
Processors	Clip, Fold, Smooth, WetDry, Wrap
Processors: Delays	AllpassDelay, CombDelay, OneTapDelay, Stutter
Processors: Distortion	Resample, SampleAndHold, Squiz, WaveShaper
Processors: Dynamics	Compressor, Gate, Maximiser, RMS
Processors: Filters	BiquadFilter, DCFilter, EQ, MoogVCF, SVFilter
Processors: Panning	AzimuthPanner, ChannelPanner, SpatialPanner, StereoBalance, StereoPanner, StereoWidth
Sequencing	ClockDivider, Counter, Euclidean, FlipFlop, ImpulseSequence, Index, Latch, Sequence
Stochastic	Logistic, PinkNoise, RandomBrownian, RandomChoice, RandomCoin, RandomExponentialDist, RandomExponential, RandomGaussian, RandomImpulseSequence, RandomImpulse, RandomUniform, StochasticNode, WhiteNoise

Contributors

Thanks to the following contributors:

Greg White (@gregwht): Examples, documentation and testing
Arthur Carabott (@acarabott), Tim Murray-Browne (@timmb): Windows support
Dan Stowell (@danstowell): Build improvements
Marc Fargas (@telenieko): Generous donation of the signalflow pypi namespace

signalflow's People

Contributors

Stargazers

Watchers

Forkers

tomad cequencer gypsybud eljeff timmb luomor-ai danstowell zeos mingyangsil acarabott assertfail victor-shepardson chriskiefer eloimarin

signalflow's Issues

Further examples

Incorporate JSON parser library

Support for loading and parsing JSON graphs, probably using json11

Node: Multi-channel panner

Multi-channel panner (2D).

Incorporate unit test framework

CUnit?

Rename channels_out → num_output_channels for consistency

Following pattern of num_things as used in other fields.

Create dedicated OSC server

Support for commands and configuration via OpenSoundControl.

Synth: Ability to stop playback and remove from graph

Add node-constant operations where node is the RHS operand

Currently, node * 0.1 is valid but 0.1 * node throws an error.

Export Patch to json

SynthDef: Support for named inputs and outputs

A SynthDef should allow naming of inputs and outputs so that synths can be chained together and parametrised properly.

Update basic nodes to handle multi-channel inputs

Sine (etc) naively assume a mono input. Extend to support multichannel.

Python interface

With https://github.com/pybind/pybind11/

Node: Linear scale

Naming review

Review variable, class and function naming for consistency and clarity.

Introduce configuration file for runtime parameters

Probably json format, possibly gengetopt.

Node: Triangle

Triangular wave

Prevent a node from being stepped multiple times per tick

Connecting a node to two outputs causes its next method to be called twice.

SynthRef

Support for reference-counted Synth objects.

Buffer: Interpolation modes

Buffers or Buffer-reading devices (Sampler, Wavetable, Waveshaper) should be configurable with interpolation modes (none, linear, cubic).

Support for a `Node::init` method

Some units cannot initialise their values until they have been wired into the graph (see signum/rnd/noise.cpp). Implement an init method across the board which allows input and output units to be queried.

Clearer interfaces, with more rigorous use of private/protected fields

Fields such as out should be declared as private with public accessors.

SynthRegistry

Add global SynthRegistry object to index Structures by name

Correct loading of multichannel buffers

Buffers with >1 channel are being reduced to mono.

Vamp support for live audio feature extraction

Node: Compressor

Compression with sidechain support

Node: HPF

Basic filters.

Optimization review

Review bottlenecks and locations to optimise, particularly with respect to vectorisation and inlining.

Node: MoogVCF

Filter.

Non-realtime processing

Offline DSP that is not driven by the audio I/O layer, probably based on a dummy device. Useful for batch generation and processing, and rendering audio that is too complex for real-time.

Vamp: Test and fix for non-standard block sizes

Node: Waveshaper

Buffer recording

Support for recording into Buffer objects. Will need some attention to where and when graph->pull_input is called (currently requires that all units are eventually connected to audioout).

Node: EQ

3-channel equaliser

Support for non-44k samplerates

Untested, but unlikely to work. Needs better handling of sample rates throughout (with distinct rates for input and output), dictated by the audio I/O layer.