pedalboard 0.5.5

pedalboard is a Python library for manipulating audio: adding effects, reading, writing, and more. It supports a number of common audio effects out of the box, and also allows the use of VST3® and Audio Unit plugin formats for third-party effects. It was built by Spotify's Audio Intelligence Lab to enable using studio-quality audio effects from within Python and TensorFlow.

Internally at Spotify, pedalboard is used for data augmentation to improve machine learning models. pedalboard also helps in the process of content creation, making it possible to add effects to audio without using a Digital Audio Workstation.

Features

• Built-in support for a number of basic audio transformations, including:
  • Guitar-style effects: Chorus, Distortion, Phaser
  • Loudness and dynamic range effects: Compressor, Gain, Limiter
  • Equalizers and filters: HighpassFilter, LadderFilter, LowpassFilter
  • Spatial effects: Convolution, Delay, Reverb
  • Pitch effects: PitchShift
  • Lossy compression: GSMFullRateCompressor, MP3Compressor
  • Quality reduction: Resample, Bitcrush
• Supports VST3® plugins on macOS, Windows, and Linux (pedalboard.load_plugin)
• Supports Audio Units on macOS
• Built-in audio I/O utilities (pedalboard.io.AudioFile)
  • Support for reading and writing AIFF, FLAC, MP3, OGG, and WAV files on all platforms with no dependencies
  • Additional support for reading AAC, AC3, WMA, and other formats depending on platform
• Strong thread-safety, memory usage, and speed guarantees
  • Releases Python's Global Interpreter Lock (GIL) to allow use of multiple CPU cores
    • No need to use multiprocessing!
  • Even when only using one thread:
    • Processes audio up to 300x faster than pySoX for single transforms, and 2-5x faster¹ than SoxBindings
    • Reads audio files up to 4x faster than librosa.load (in many cases)
• Tested compatibility with TensorFlow - can be used in tf.data pipelines!

Installation

pedalboard is available via PyPI (via Platform Wheels):

pip install pedalboard

If you are new to Python, follow INSTALLATION.md for a robust guide.

Compatibility

pedalboard is thoroughly tested with Python 3.6, 3.7, 3.8, 3.9, and 3.10 as well as experimental support for Python 3.11 and PyPy 3.7, 3.8, and 3.9.

• Linux
  • Tested heavily in production use cases at Spotify
  • Tested automatically on GitHub with VSTs
  • Platform manylinux wheels built for x86_64 (Intel/AMD) and aarch64 (ARM/Apple Silicon)
  • Most Linux VSTs require a relatively modern Linux installation (with glibc > 2.27)
• macOS
  • Tested manually with VSTs and Audio Units
  • Tested automatically on GitHub with VSTs
  • Platform wheels available for both Intel and Apple Silicon
  • Compatible with a wide range of VSTs and Audio Units
• Windows
  • Tested automatically on GitHub with VSTs
  • Platform wheels available for amd64 (x86-64, Intel/AMD)

Plugin Compatibility

pedalboard allows loading VST3® and Audio Unit plugins, which could contain any code. Most plugins that have been tested work just fine with pedalboard, but some plugins may not work with pedalboard; at worst, some may even crash the Python interpreter without warning and with no ability to catch the error. For an ever-growing compatibility list, see COMPATIBILITY.md.

Most audio plugins are "well-behaved" and conform to a set of conventions for how audio plugins are supposed to work, but many do not conform to the VST3® or Audio Unit specifications. pedalboard attempts to detect some common programming errors in plugins and can work around many issues, including automatically detecting plugins that don't clear their internal state when asked. Even so, plugins can misbehave without pedalboard noticing.

If audio is being rendered incorrectly or if audio is "leaking" from one process() call to the next in an undesired fashion, try:

1. Passing silence to the plugin in between calls to process(), to ensure that any reverb tails or other internal state has time to fade to silence
2. Reloading the plugin every time audio is processed (with pedalboard.load_plugin)

Examples

Quick Start

from pedalboard import Pedalboard, Chorus, Reverb
from pedalboard.io import AudioFile

# Read in a whole audio file:
with AudioFile('some-file.wav', 'r') as f:
  audio = f.read(f.frames)
  samplerate = f.samplerate

# Make a Pedalboard object, containing multiple plugins:
board = Pedalboard([Chorus(), Reverb(room_size=0.25)])

# Run the audio through this pedalboard!
effected = board(audio, samplerate)

# Write the audio back as a wav file:
with AudioFile('processed-output.wav', 'w', samplerate, effected.shape[0]) as f:
  f.write(effected)

Making a guitar-style pedalboard

# Don't do import *! (It just makes this example smaller)
from pedalboard import *
from pedalboard.io import AudioFile

with AudioFile('guitar-input.wav', 'r') as f:
  audio = f.read(f.frames)
  samplerate = f.samplerate

# Make a pretty interesting sounding guitar pedalboard:
board = Pedalboard([
    Compressor(threshold_db=-50, ratio=25),
    Gain(gain_db=30),
    Chorus(),
    LadderFilter(mode=LadderFilter.Mode.HPF12, cutoff_hz=900),
    Phaser(),
    Convolution("./guitar_amp.wav", 1.0),
    Reverb(room_size=0.25),
])

# Pedalboard objects behave like lists, so you can add plugins:
board.append(Compressor(threshold_db=-25, ratio=10))
board.append(Gain(gain_db=10))
board.append(Limiter())

# ... or change parameters easily:
board[0].threshold_db = -40

# Run the audio through this pedalboard!
effected = board(audio, samplerate)

# Write the audio back as a wav file:
with AudioFile('processed-output.wav', 'w', samplerate, effected.shape[0]) as f:
  f.write(effected)

Using VST3® or Audio Unit plugins

from pedalboard import Pedalboard, Reverb, load_plugin
from pedalboard.io import AudioFile

# Load a VST3 or Audio Unit plugin from a known path on disk:
vst = load_plugin("./VSTs/RoughRider3.vst3")

print(vst.parameters.keys())
# dict_keys([
#   'sc_hpf_hz', 'input_lvl_db', 'sensitivity_db',
#   'ratio', 'attack_ms', 'release_ms', 'makeup_db',
#   'mix', 'output_lvl_db', 'sc_active',
#   'full_bandwidth', 'bypass', 'program',
# ])

# Set the "ratio" parameter to 15
vst.ratio = 15

# Use this VST to process some audio:
with AudioFile('some-file.wav', 'r') as f:
  audio = f.read(f.frames)
  samplerate = f.samplerate
effected = vst(audio, samplerate)

# ...or put this VST into a chain with other plugins:
board = Pedalboard([vst, Reverb()])
# ...and run that pedalboard with the same VST instance!
effected = board(audio, samplerate)

Creating parallel effects chains

This example creates a delayed pitch-shift effect by running multiple Pedalboards in parallel on the same audio. Pedalboard objects are themselves Plugin objects, so you can nest them as much as you like:

from pedalboard import Pedalboard, Compressor, Delay, Distortion, Gain, PitchShift, Reverb, Mix

passthrough = Gain(gain_db=0)

delay_and_pitch_shift = Pedalboard([
  Delay(delay_seconds=0.25, mix=1.0),
  PitchShift(semitones=7),
  Gain(gain_db=-3),
])

delay_longer_and_more_pitch_shift = Pedalboard([
  Delay(delay_seconds=0.5, mix=1.0),
  PitchShift(semitones=12),
  Gain(gain_db=-6),
])

board = Pedalboard([
  # Put a compressor at the front of the chain:
  Compressor(),
  # Run all of these pedalboards simultaneously with the Mix plugin:
  Mix([
    passthrough,
    delay_and_pitch_shift,
    delay_longer_and_more_pitch_shift,
  ]),
  # Add a reverb on the final mix:
  Reverb()
])

For more examples, see:

• the examples folder of this repository
• the Pedalboard Demo Colab notebook
• an interactive web demo on Hugging Face Spaces and Gradio (via @AK391)

Contributing

Contributions to pedalboard are welcomed! See CONTRIBUTING.md for details.

Frequently Asked Questions

Can Pedalboard be used with live (real-time) audio?

Technically, yes, Pedalboard could be used with live audio input/output. See:

• Pull request #98, which contains an experimental live audio interface written in C++.
• @stefanobazzi's guitarboard project for an example that uses the python-sounddevice library.

However, there are a couple big caveats when talking about using Pedalboard in a live context. Python, as a language, is garbage-collected, meaning that your code randomly pauses on a regular interval to clean up unused objects. In most programs, this is not an issue at all. However, for live audio, garbage collection can result in random pops, clicks, or audio drop-outs that are very difficult to prevent. Python's Global Interpreter Lock also adds potentially-unbounded delays when switching threads, and most operating systems use a separate high-priority thread for audio processing; meaning that Python could block this thread and cause stuttering if any Python objects are accessed or mutated in the audio thread.

Note that if your application processes audio in a streaming fashion, but allows for large buffer sizes (multiple seconds of audio) or soft real-time requirements, Pedalboard can be used there without issue. Examples of this use case include streaming audio processing over the network, or processing data offline but chunk-by-chunk.

Does Pedalboard support changing a plugin's parameters over time?

Yes! While there's no built-in function for this, it is possible to vary the parameters of a plugin over time manually:

from pedalboard.io import AudioFile
from pedalboard import Pedalboard, Compressor, Reverb
from tqdm import tqdm

board = Pedalboard([Compressor(), Reverb()])
reverb = board[1]

# Smaller step sizes would give a smoother transition,
# at the expense of processing speed
step_size_in_samples = 100

# Manually step through the audio _n_ samples at a time, reading in chunks:
with AudioFile("sample.wav") as af:

    # Open the output audio file so that we can directly write audio as we process, saving memory:
    with AudioFile(
        "sample-processed-output.wav", "w", af.samplerate, num_channels=af.num_channels
    ) as o:

        # Create a progress bar to show processing speed in real-time:
        with tqdm(total=af.frames, unit=' samples') as pbar:
            for i in range(0, af.frames, step_size_in_samples):
                chunk = af.read(step_size_in_samples)

                # Set the reverb's "wet" parameter to be equal to the
                # percentage through the track (i.e.: make a ramp from 0% to 100%)
                percentage_through_track = i / af.frames
                reverb.wet_level = percentage_through_track

                # Update our progress bar with the number of samples received:
                pbar.update(chunk.shape[1])

                # Process this chunk of audio, setting `reset` to `False`
                # to ensure that reverb tails aren't cut off
                output = board.process(chunk, af.samplerate, reset=False)
                o.write(output)

With this technique, it's possible to automate any parameter. Usually, using a step size of somewhere between 100 and 1,000 (2ms to 22ms at a 44.1kHz sample rate) is small enough to avoid hearing any audio artifacts, but big enough to avoid slowing down the code dramatically.

Can Pedalboard be used with VST instruments, instead of effects?

Not yet! The underlying framework (JUCE) supports VST and AU instruments just fine, but Pedalboard itself would have to be modified to support instruments.

Can Pedalboard plugins accept MIDI?

Not yet, either - although the underlying framework (JUCE) supports passing MIDI to plugins, so this would also be possible to add.

License

pedalboard is Copyright 2021-2022 Spotify AB.

pedalboard is licensed under the GNU General Public License v3. pedalboard includes a number of libraries that are statically compiled, and which carry the following licenses:

• The core audio processing code is pulled from JUCE 6, which is dual-licensed under a commercial license and the GPLv3.
• The VST3 SDK, bundled with JUCE, is owned by Steinberg® Media Technologies GmbH and licensed under the GPLv3.
• The PitchShift plugin uses the Rubber Band Library, which is dual-licensed under a commercial license and the GPLv2 (or newer).
• The MP3Compressor plugin uses libmp3lame from the LAME project, which is licensed under the LGPLv2 and upgraded to the GPLv3 for inclusion in this project (as permitted by the LGPLv2).
• The GSMFullRateCompressor plugin uses libgsm, which is licensed under the ISC license and compatible with the GPLv3.

VST is a registered trademark of Steinberg Media Technologies GmbH.