pedalboard
v0.9.16
pedalboard is a Python library for working with audio: reading, writing, rendering, adding effects, and more. It supports most popular audio file formats and a number of common audio effects out of the box, and also allows the use of VST3® and Audio Unit formats for loading third-party software instruments and effects.
pedalboard 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 and to help power features like Spotify's AI DJ and AI Voice Translation. pedalboard also helps in the process of content creation, making it possible to add effects to audio without using a Digital Audio Workstation.
O(1) memory usageAudioStreamChorus, Distortion, Phaser, ClippingCompressor, Gain, LimiterHighpassFilter, LadderFilter, LowpassFilterConvolution, Delay, ReverbPitchShiftGSMFullRateCompressor, MP3CompressorResample, Bitcrushpedalboard.load_plugin)multiprocessing!tf.data pipelines!pedalboard is available via PyPI (via Platform Wheels):
pip install pedalboard # That's it! No other dependencies required.
If you are new to Python, follow INSTALLATION.md for a robust guide.
pedalboard is thoroughly tested with Python 3.8, 3.9, 3.10, 3.11, 3.12, and 3.13.
manylinux and musllinux wheels built for x86_64 (Intel/AMD) and aarch64 (ARM/Apple Silicon)amd64 (x86-64, Intel/AMD)Note: If you'd rather watch a video instead of reading examples or documentation, watch Working with Audio in Python (feat. Pedalboard) on YouTube.
from pedalboard import Pedalboard, Chorus, Reverb
from pedalboard.io import AudioFile
# Make a Pedalboard object, containing multiple audio plugins:
board = Pedalboard([Chorus(), Reverb(room_size=0.25)])
# Open an audio file for reading, just like a regular file:
with AudioFile('some-file.wav') as f:
# Open an audio file to write to:
with AudioFile('output.wav', 'w', f.samplerate, f.num_channels) as o:
# Read one second of audio at a time, until the file is empty:
while f.tell() < f.frames:
chunk = f.read(f.samplerate)
# Run the audio through our pedalboard:
effected = board(chunk, f.samplerate, reset=False)
# Write the output to our output file:
o.write(effected)Note: For more information about how to process audio through Pedalboard plugins, including how the
resetparameter works, see the documentation forpedalboard.Plugin.process.
# Don't do import *! (It just makes this example smaller)
from pedalboard import *
from pedalboard.io import AudioFile
# Read in a whole file, resampling to our desired sample rate:
samplerate = 44100.0
with AudioFile('guitar-input.wav').resampled_to(samplerate) as f:
audio = f.read(f.frames)
# 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)from pedalboard import Pedalboard, Reverb, load_plugin
from pedalboard.io import AudioFile
from mido import Message # not part of Pedalboard, but convenient!
# Load a VST3 or Audio Unit plugin from a known path on disk:
instrument = load_plugin("./VSTs/Magical8BitPlug2.vst3")
effect = load_plugin("./VSTs/RoughRider3.vst3")
print(effect.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
effect.ratio = 15
# Render some audio by passing MIDI to an instrument:
sample_rate = 44100
audio = instrument(
[Message("note_on", note=60), Message("note_off", note=60, time=5)],
duration=5, # seconds
sample_rate=sample_rate,
)
# Apply effects to this audio:
effected = effect(audio, sample_rate)
# ...or put the effect into a chain with other plugins:
board = Pedalboard([effect, Reverb()])
# ...and run that pedalboard with the same VST instance!
effected = board(audio, sample_rate)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()
])pedalboard supports streaming live audio through
an AudioStream object,
allowing for real-time manipulation of audio by adding effects in Python.
from pedalboard import Pedalboard, Chorus, Compressor, Delay, Gain, Reverb, Phaser
from pedalboard.io import AudioStream
# Open up an audio stream:
with AudioStream(
input_device_name="Apogee Jam+", # Guitar interface
output_device_name="MacBook Pro Speakers"
) as stream:
# Audio is now streaming through this pedalboard and out of your speakers!
stream.plugins = Pedalboard([
Compressor(threshold_db=-50, ratio=25),
Gain(gain_db=30),
Chorus(),
Phaser(),
Convolution("./guitar_amp.wav", 1.0),
Reverb(room_size=0.25),
])
input("Press enter to stop streaming...")
# The live AudioStream is now closed, and audio has stopped.tf.data Pipelinesimport tensorflow as tf
sr = 48000
# Put whatever plugins you like in here:
plugins = pedalboard.Pedalboard([pedalboard.Gain(), pedalboard.Reverb()])
# Make a dataset containing random noise:
# NOTE: for real training, here's where you'd want to load your audio somehow:
ds = tf.data.Dataset.from_tensor_slices([np.random.rand(sr)])
# Apply our Pedalboard instance to the tf.data Pipeline:
ds = ds.map(lambda audio: tf.numpy_function(plugins.process, [audio, sr], tf.float32))
# Create and train a (dummy) ML model on this audio:
model = tf.keras.models.Sequential([tf.keras.layers.InputLayer(input_shape=(sr,)), tf.keras.layers.Dense(1)])
model.compile(loss="mse")
model.fit(ds.map(lambda effected: (effected, 1)).batch(1), epochs=10)For more examples, see:
Contributions to pedalboard are welcomed! See CONTRIBUTING.md for details.
To cite pedalboard in academic work, use its entry on Zenodo:
To cite via BibTeX:
@software{sobot_peter_2023_7817838,
author = {Sobot, Peter},
title = {Pedalboard},
month = jul,
year = 2021,
publisher = {Zenodo},
doi = {10.5281/zenodo.7817838},
url = {https://doi.org/10.5281/zenodo.7817838}
}pedalboard is Copyright 2021-2024 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:
PitchShift plugin and time_stretch functions use the Rubber Band Library, which is dual-licensed under a commercial license and the GPLv2 (or newer). FFTW is also included to speed up Rubber Band, and is licensed under the GPLv2 (or newer).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).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.
