#microtonal #midi #modular #piano #synthesizer

app microwave

Make xenharmonic music and explore musical tunings

34 breaking releases

0.35.0 May 12, 2023
0.34.0 Feb 22, 2023
0.33.0 Dec 10, 2022
0.32.0 Nov 10, 2022
0.7.1 Jul 26, 2020

#20 in Audio

Download history 55/week @ 2023-05-29 1/week @ 2023-06-05 2/week @ 2023-06-12 53/week @ 2023-06-26 30/week @ 2023-07-03 1/week @ 2023-07-10 2/week @ 2023-07-24 2/week @ 2023-07-31 2/week @ 2023-08-07 40/week @ 2023-08-14 238/week @ 2023-08-21 70/week @ 2023-08-28 45/week @ 2023-09-04 47/week @ 2023-09-11

400 downloads per month

MIT and LGPL-2.1

4MB
17K SLoC

Make xenharmonic music and explore musical tunings.

Resources

Overview

microwave is a microtonal modular waveform synthesizer and effects processor with soundfont rendering capabilities based on:

  • tune – a microtonal library
  • magnetron – a modular synthesizer architecture
  • fluid-xenth – a microtonal soundfont renderer
  • Bevy – a graphics and game engine

It features a virtual piano UI enabling you to play polyphonic microtonal melodies with your touch screen, computer keyboard, MIDI keyboard or mouse. The UI provides information about pitches and just intervals in custom tuning systems.

The built-in modular synthesis engine does not use any fixed architecture and can be customized to react to all sorts of input events.

Demo

Download / Installation

Option A: Try out the web app to get a very first impression:

Option B: Download a precompiled version of microwave for the supported target architectures:

Option C: Use Cargo to build a fresh binary from scratch for your own target architecture:

# If you are using Linux: Make sure all dev dependencies are installed.
# On the CI environment (Ubuntu 20.04) we only need to add two dev dependencies:
sudo apt install libasound2-dev libudev-dev

# Make sure pkg-config is installed
sudo apt install pkg-config

cargo install -f microwave

microwave should run out-of-the box on a recent (Ubuntu) Linux, Windows or macOS installation. If it doesn't, the problem is probably caused by the Bevy framework. In that case, try following these instructions.

Usage

Hint: Run microwave with parameters from a shell environment (Bash, PowerShell). Double-clicking on the executable will not provide access to all features!

microwave run                       # 12-EDO scale (default)
microwave run steps 1:22:2          # 22-EDO scale
microwave run scl-file my_scale.scl # imported scale
microwave run help                  # Print help about how to set the parameters to start microwave

This should spawn a window displaying a virtual keyboard. Use your touch screen, computer keyboard or mouse to play melodies on the virtual piano.

Profiles

On startup, microwave tries to load a profile specified by the -p / --profile parameter or the MICROWAVE_PROFILE environment variable. If no such file is found microwave will create a default profile for you.

Example Profiles

microwave is shipped with the following example profiles:

  • audio-effect.yml: Demo showing how to configure an effect-only profile.
  • microwave.yml: The default profile created at first startup.
  • sympathetic.yml: Demo showing how to use note-input controlled waveguides to achieve a sympathetic resonance effect.

To use a profile, run:

microwave -p <profile-name>

Profile Structure

The profile has the following structure:

num_buffers:        # Number of main audio buffers
audio_buffers:      # Audio buffers that are played back by the main audio device
waveform_templates: # Named templates to be used by the Magnetron synthesizer
waveform_envelopes: # Named envelopes to be used by the Magnetron synthesizer
effect_templates:   # Named templates to be used by the effect processors
stages:             # Stages that can create or process audio or MIDI data

LF (Low-Frequency) Sources

Almost all numerical profile parameters can update in real-time. To keep the audio engine performant updates are usually evaluated at a much lower rate than the audio sampling rate. LF sources, therefore, add control and expressiveness to your playing but aren't well suited for spectral sound modulation.

To define an LF source the following data types can be used:

  • Numbers, e.g.
    frequency: 440.0
    
  • Strings referring to a named template, e.g.
    frequency: WaveformPitch
    
  • Nested LF source expressions, e.g.
    frequency: { Mul: [ 2.0, WaveformPitch ] }
    
    or (using indented style)
    frequency:
      Mul:
        - 2.0
        - WaveformPitch
    

Unfortunately, no detailed LF source documentation is available yet. However, the example profile, microwave's error messages and basic YAML knowledge should enable you to find valid LF source expressions.

waveform_templates Section

The purpose of the waveform_templates section of the profile is to define the most important LF sources s.t. they do not have to be redefined over and over again. The default profile contains some templates that will be explained in the following paragraphs.

WaveformPitch and WaveformPeriod Templates

waveform_templates:
  - name: WaveformPitch
    value:
      Mul:
        - Property: WaveformPitch
        - Semitones:
            Controller:
              kind: PitchBend
              map0: 0.0
              map1: 2.0
  - name: WaveformPeriod
    value:
      Mul:
        - Property: WaveformPeriod
        - Semitones:
            Controller:
              kind: PitchBend
              map0: 0.0
              map1: -2.0

The given fragment defines a template with name WaveformPitch or WaveformPeriod, respectively. The output values are calculated by reading the waveform's WaveformPitch/WaveformPeriod property and multiplying it with the pitch-bend wheel's value in whole tones.

Note: Reacting to pitch-bend events is not a hardcoded feature of microwave but a behavior that the user can define by themself!

Fadeout Template

waveform_templates:
  - name: Fadeout
    value:
      Controller:
        kind: Damper
        map0: { Property: OffVelocitySet }
        map1: 0.0

The Fadeout template provides a value describing to what extent a waveform is supposed to be faded out. It works in the following way:

  • While a key is pressed down, OffVelocitySet resolves to 0.0. As a result, Controller, as well, resolves to 0.0, regardless of the damper pedal state. Therefore, the waveform remains stable.
  • As soon as a key is released, OffVelocitySet will resolve to 1.0. Now, Controller will interpolate between 1.0 (map0 = damper released released) and 0.0 (map1 = damper pedal pressed). As a consequence, the waveform will fade out unless the damper pedal is pressed.

Note: Like in the examples before, reacting to the damper pedal is not a hardcoded feature built into microwave but customizable behavior.

EnvelopeL and EnvelopeR Templates

waveform_templates:
  - name: EnvelopeL
    value:
      Mul:
        - Controller:
            kind: Pan
            map0: { Property: Velocity }
            map1: 0.0
        - Controller:
            kind: Volume
            map0: 0.0
            map1: 0.25
  - name: EnvelopeR
    value:
      Mul:
        - Controller:
            kind: Pan
            map0: 0.0
            map1: { Property: Velocity }
        - Controller:
            kind: Volume
            map0: 0.0
            map1: 0.25

These templates are designed to provide a reasonable envelope amplitude of ≈ -18dB which is sensitive to the pan controller, the volume controller and the pressed key's velocity. The result is obtained by multiplying the following quantities:

  • Pan controller: From 1 to 0 (left channel) or 0 to 1 (right channel) (≈ -6dB)
  • Velocity controller: From 0 to 0.25 (≈ -12dB)
  • Key velocity: From 0 to 1

Note: You are not forced to couple envelope amplitudes to those quantities. For example, you could replace the pan controller with the balance controller. Use an LF source that matches your use case.

How to Access the Templates

Just provide the name of the template as a single string argument. Examples:

frequency: WaveformPitch
fadeout: Fadeout
out_levels: [EnvelopeL, EnvelopeR]

waveform_envelopes Section

Every waveform needs to refer to an envelope defined in the waveform_envelopes section of the config file. Envelopes transfer the result of the internal waveform buffers to the main audio pipeline and limit the waveform's lifetime.

An envelope definition typically looks like the following:

waveform_envelopes:
  - name: Piano
    fadeout: Fadeout
    attack_time: 0.01
    decay_rate: 1.0
    release_time: 0.25
    in_buffer: 7
    out_buffers: [0, 1]
    out_levels: [EnvelopeL, EnvelopeR]

with

  • name: The name of the envelope.
  • fadeout: The amount by which the waveform should fade out. Important: If this value is set to constant 0.0 the waveform will never fade out and continue to consume CPU resources, eventually leading to an overload of the audio thread.
  • attack_time: The linear attack time in seconds.
  • decay_rate: The exponential decay rate in 1/seconds (inverse half-life) after the attack phase is over.
  • release_time: The linear release time in seconds. The waveform is considered exhausted as soon as the integral over fadeout / release_time * dt reaches 1.0.
  • in_buffer: The waveform buffer containing the signal that is supposed to be enveloped.
  • out_buffers: The (stereo) buffers of the main audio pipeline that the enveloped signal is supposed to be written to.
  • out_levels: The amplitude factor to apply when writing to the main audio pipeline. It makes sense to use EnvelopeL/EnvelopeR as a value but the user can choose whatever LF source expression they find useful.

effect_templates Section

This section is completely analogous to the waveform_templates section but it is dedicated to work in combination with the Effect stages documented below. One key difference is that it is not able to access waveform-specific properties like Velocity, KeyPressure, etc.

stages Section / Main Audio Pipeline

The stages section defines the sections that are evaluated sequentially while the main audio thread is running. Not all sections (e.g. MidiOut) contribute to the main audio pipeline but it makes sense to declare them here anyway. Some of the stages, the output targets, are sensitive to note inputs. In that case, the note_input property has to be set. It can have the following values:

  • Foreground: Only activate notes when the given output target is currently active.
  • Background: Always activate notes when a note event is received.

Magnetron Synthesizer – Create Your Own Waveforms

To enable the modular magnetron synthesizer engine add the following stage:

stages:
  - Magnetron:
      note_input: Foreground
      num_buffers: # Number of waveform audio buffers
      waveforms:   # Waveform definitions

waveforms Section

The waveforms section defines the waveform render stages to be applied sequentially when a waveform is triggered.

You can mix and match as many stages as you want to create the tailored sound you wish for. The following example config defines a clavinettish sounding waveform that I discovered by accident:

waveforms:
  - name: Funky Clavinet
    envelope: Piano
    stages:
      - Oscillator:
          kind: Sin
          frequency: WaveformPitch
          modulation: None
          out_buffer: 0
          out_level: 440.0
      - Oscillator:
          kind: Triangle
          frequency: WaveformPitch
          modulation: ByFrequency
          mod_buffer: 0
          out_buffer: 1
          out_level: 1.0
      - Filter:
          kind: HighPass2
          resonance:
            Mul:
              - WaveformPitch
              - Fader:
                  movement: 10.0
                  map0: 2.0
                  map1: 4.0
          quality: 5.0
          in_buffer: 1
          out_buffer: 7
          out_level: 1.0

While rendering the sound the following stages are applied:

  1. Generate a sine wave with the waveform's nominal frequency F and an amplitude of 440. Write this waveform to buffer 0.
  2. Generate a triangle wave with frequency F and an amplitude of 1.0. Modulate the waveform's frequency (in Hz) sample-wise by the amount stored in buffer 0. Write the modulated waveform to buffer 1.
  3. Apply a second-order high-pass filter to the samples stored in buffer 1. The high-pass's resonance frequency rises from 2F to 4F within 0.1 seconds. Write the result to buffer 7.
  4. Wrap an envelope around the signal in buffer 7 and transfer the enveloped signal to buffer 0 and 1 of the main audio pipeline. This is the behavior defined for the Piano envelope in the waveform_envelopes section (see above).

To create your own waveforms use the default config file as a starting point and try editing it by trial-and-error. Let microwave's error messages guide you to find valid configurations.

Fluid Synthesizer

For playback of sampled sounds you need to add a Fluid stage to the stages section.

The following example starts up a Fluid stage which renders the contents of a given soundfont file. The rendered audio will be written to the audio buffers 0 and 1 of the main audio pipeline.

stages:
  - Fluid:
      note_input: Foreground
      soundfont_location: <soundfont-location>
      out_buffers: [0, 1]

If you like to use compressed sf3 files you need to compile microwave with the sf3 feature enabled. Note that the startup will take significantly longer since the soundfont needs to be decompressed first.

Effects – Create Your Own Effects

To add your own customized effects add a Generic stage. The following config will add a rotary-speaker effect stage to the main audio pipeline.

stages:
  - Generic:
      Effect:
        RotarySpeaker:
          buffer_size: 100000
          gain:
            Controller:
              kind: Sound9
              map0: 0.0
              map1: 0.5
          rotation_radius: 20.0
          speed:
            Fader:
              movement:
                Controller:
                  kind: Sound10
                  map0: -2.0
                  map1: 1.0
              map0: 1.0
              map1: 7.0
          in_buffers: [4, 5]
          out_buffers: [14, 15]

The given config defines the following properties:

  • buffer_size: A fixed delay buffer size of 100000 samples.
  • gain: An input gain ranging from 0.0 to 0.5. The input gain can be controlled via the F9 key or MIDI CCN 78.
  • rotation_radius: A rotation radius of 20 cm.
  • speed: A rotation speed ranging from 1 Hz to 7 Hz. The selected speed is determined by the Fader component which will gradually fade between the two values. The movement of the fader is controlled by the the F10 key or MIDI CCN 79 and ranges from -2.0/s to 1.0/s in order to simulate the rotary speaker's deceleration and acceleration.
  • in_buffers: Buffers 4 and 5 are used as effect inputs.
  • out_buffers: Buffers 14 and 15 are used as effect outputs.

MIDI Out

To enable playback through an external MIDI device add the following stage to the audio pipeline:

stages:
  - MidiOut:
      note_input: Foreground
      out_device: <midi-device>
      out_channel: 0
      num_out_channels: 9
      device_id: 127
      tuning_program: 0
      tuning_method: <tuning-method>

The available tuning methods are full, full-rt, octave-1, octave-1-rt, octave-2, octave-2-rt, fine-tuning and pitch-bend.

To retrieve a list of available MIDI devices run:

microwave devices

Command Line Parameters

The command-line enables you to set set up sample rates, buffer sizes and other startup parameters. To print a full list of available command-line arguments run:

microwave run help

MIDI In

To listen for events originating from an external MIDI device you need to specify the name of the input device:

microwave devices # List MIDI devices
microwave run --midi-in name-of-my-device
microwave run --midi-in "name of my device" # If the device name contains spaces

Live Interactions

You can live-control your waveforms and effects with your mouse pointer, touch pad or any MIDI Control Change messages source.

The following example stage defines a resonating low-pass filter whose resonance frequency can be controlled with a MIDI modulation wheel/lever from 2,000 to 10,000 Hz.

Filter:
  kind: LowPass2
  resonance:
    Controller:
      kind: Modulation
      map0: 2000.0
      map1: 10000.0
  quality: 5.0
  in_buffer: 0
  out_buffer: 7
  out_level: 1.0

If you want to use the mouse's vertical axis for sound control use the Breath controller.

resonance:
  Controller:
    kind: Breath
    map0: 2000.0
    map1: 10000.0

If you want to use the touchpad for polyphonic sound control use the KeyPressure property.

resonance:
  Linear:
    input:
      Property: KeyPressure
    map0: 2000.0
    map1: 10000.0

Note: While Controller values are scaled to 0..1 (or -1..1 in the case of pitch-bend events) and require a range mapping (map0/map1 parameters), Property values can be directly digested. If necessary, they can be To rescaled using Mul or Linear.

Feature List

  • Sound features
    • Built-in modular waveform synthesizer with physical modeling synthesis
      microwave run -p <profile-location> [scale-expression]
      
    • Soundfont renderer
    • External synthesizer via MIDI-out
    • Microphone / aux input
    • WAV recording
  • Control features
    • Sequencer / piano keyboard via MIDI-in
      microwave run --midi-in <midi-source> [scale-expression]
      
    • Lumatone / multichannel input
      # 31-EDO Lumatone preset centered around D4 (62, Layout offset -5)
      microwave ref-note 62 --root 57 --luma-offs 31 --lo-key 0 --up-key 155 --midi-in lumatone steps 1:31:2
      
    • Computer keyboard (configurable isomorphic layout)
    • Touch Screen
    • Mouse
    • Channel events (pitch-bend, modulation, pedals, aftertouch, etc.)
    • Polyphonic events (key pressure)
    • LF sources (envelopes, time slices, oscillators, etc.)
  • Effects
    • Low-pass
    • 2nd order low-pass
    • High-pass
    • 2nd order high-pass
    • Band-pass
    • Notch filter
    • All-pass
    • Reverb
    • Spatial delay
    • Rotary speaker
  • Microtuning features
    • Custom scales
    • SCL imports
    • KBM imports
    • Tuning-dependent automatic isomorphic keyboard layouts
    • MIDI-out retuning via different tuning message types
    • Display frequencies and rational number approximations
    • Customizable second visual keyboard (--kb2 option)

Help

For a complete list of command line options run

microwave help

License

microwave statically links against OxiSynth for soundfont rendering capabilities. This makes the binary executable of microwave a derivative work of OxiSynth. OxiSynth is licensed under the GNU Lesser General Public License, version 2.1.

Dependencies

~27–72MB
~1M SLoC