Skip to main content

Simple wrapper for librubberband

Project description


rubberband is a simple Python3 wrapper around the well-known librubberband sound stretching / pitch-shifting library. Unlike existing Python wrappers (e.g. pyrubberband) this is a true native extension.

The initial release provides a single function that will stretch a mono audio stream by multiplying its duration by a provided factor. Future versions may include pitch shifting, and more complex processing based on data maps.


The module is available only for MacOS and Linux. The code may compile on Windows, but it has not been tested. Dependencies are:

Assuming these dependencies are met, then installation is simplicity itself:

pip3 install rubberband

The install script does check for the required libraries, and will complain vociferously if they cannot be located. Information on obtaining them is available from the links above.


Audio stream formatting

The module exposes a single function rubberband.stretch which applies the librubberband algorithm to a mono audio stream, encoded in one of the following formats, each of which has a corresponding constant, as set out in the third column:

Format Description Constant
PCM_U8 unsigned 8-bit rubberband.uint8
PCM_S8 signed 8-bit rubberband.int8
PCM_16 signed 16-bit rubberband.int16
PCM_32 signed 32-bit rubberband.int32
FLOAT normalised 32-bit float rubberband.float32

Note that floating point data is assumed to be normalised, so all samples lie in the range [-1,1).

Audio data can be passed to rubberband.stretch in any of three ways:

Typed array

A 1-dimensional NUMPY typed array object, whose dtype is one of numpy.uint8, numpy.int8, numpy.int16, numpy.int32 or numpy.float32. The type of the audio data is deduced from this, using the strangely convenient fact that if T is one of uint8, int8, int16, int32, float32 then

numpy.dtype(numpy.T).num == rubberband.T
A simple Python list, all of whose elements are of a type implicitly convertible to float. In this case, the audio format cannot be deduced, so it must be specified using the format argument to rubberband.stretch (see below).
Raw bytestream
A Python bytes object, whose content is the raw PCM byte stream (note: audio file metadata, e.g. WAV file headers, must be stripped, so only PCM data remains). Again, in this case, the audio format cannot be deduced, so it must be specified using the format argument to rubberband.stretch (see below).

In all cases, the output from rubberband.stretch has the same PCM format, and is stored in the same kind of object, as the input. So, for example, given a bytes object representing a PCM_16 audio channel, rubberband.stretch returns bytes object representing a PCM_16 stretched audio channel.

Method signature

rubberband.stretch (input, format = rubberband.float32, rate = 48000 , ratio = 1 , crispness = 5 , formants = False, precise = True )


The input is assumed to represent a single channel of PCM audio data, encoded with one of the schemes listed above. It can be any of the types set out above.
The PCM format of the data, specified using one of the constants set out above. This value is ignored if input is a NUMPY typed array, in which case the format is deduced from its dtype.
The frame rate of the input audio stream (so bit rate divided by sample size).
The ratio of output length to input length (in seconds / number of samples).
Integer 0 - 6, default 5: measure of performance - see the rubberband-cli documentation for more details.
Boolean, default False : whether or not to preserve formants - see the rubberband-cli documentation for more details.
Boolean, default True : whether or not to use the precise stretching algorithm - see the rubberband-cli documentation for more details.
Return value
An object containing the stretched audio data, represented using the same PCM encoding as the input. Samples are normalised to lie in the expected range for the format.


import rubberband
import soundfile

print(f'Raw input type is : {type(data)}')



Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for rubberband, version 1.0.2
Filename, size File type Python version Upload date Hashes
Filename, size rubberband-1.0.2.tar.gz (99.4 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page