Skip to content

Latest commit

 

History

History
262 lines (197 loc) · 7.9 KB

README.md

File metadata and controls

262 lines (197 loc) · 7.9 KB
Raw

python-wavefile

CI Coverage Status PyPi Downloads PyPi Version

Pythonic libsndfile wrapper to read and write audio files.

Features

  • Wave file resources open and close as context managers (with)
  • Property accessors for format, channels, length, sample rate... and metadata (ID3...)
  • Real multichannel (not just mono/stereo, but also surround, ambisonics and virtually any number of channels)
  • All libsndfile formats supported, floating point encodings used by default
  • Numpy based interface
  • Generators for efficient block-by-block access
  • Alternative Matlab-like whole-file interface. Less efficient, but convenient for quick and dirty scripts.
  • Shortened constant names for formats (Using namespaces instead of prefixes)
  • Tools for listing available formats
  • Transparent Unicode handling for filenames and text strings
  • No module compilation required (wraps the dll using ctypes)
  • Compatible with Python >= 3.8

You can find the latest version at: https://github.com/vokimon/python-wavefile

Installation

Latest Version Supported Python Versions

Binary dependencies

Python dependencies are managed by the setup.py script. But still there are a couple of binary dependencies. In Debian/Ubuntu, you can install them by casting:

sudo apt-get install -y libsndfile1 portaudio19-dev

PortAudio and its Python wrapper, PyAudio, are just required in order to run the examples.

Using PyPi

pip install wavefile

From sources

If you want to develop, installing editable is recomended. From the source directory:

pip install -e .

Examples

Whole file (slow) processing

Writing and reading whole audio files (like Mathlab primitives do) is not as efficient than block based processing to process long audios. But it could be quite convenient on occassions to have quickly something working.

import wavefile
import numpy as np

def sinusoid(samples, f, samplerate=44100):
    return np.sin( np.linspace(0, 2*np.pi*f*samples/samplerate, samples))[:,np.newaxis]

def channels(*args):
    return np.hstack(args).T


audio = channels(
    sinusoid(100000,  440),
    sinusoid(100000,  880),
    sinusoid(100000, 1760),
)

wavefile.save("sinusoid.wav", audio, 44100)

loadedsamplerate, loaded = wavefile.load("sinusoid.wav")

loaded.shape() # 3, 100000

Block writing example

from wavefile import WaveWriter, Format
import numpy as np

BUFFERSIZE = 512
NCHANNELS = 2

with WaveWriter(
    'synth.ogg',
    channels=NCHANNELS,
    format=Format.OGG|Format.VORBIS,
) as w:
    w.metadata.title = "Some Noise"
    w.metadata.artist = "The Artists"
    data = np.zeros((NCHANNELS,BUFFERSIZE), np.float32)
    for x in range(256):
        # First channel: Saw wave sweep
        data[0,:] = (x*np.arange(BUFFERSIZE, dtype=np.float32)%BUFFERSIZE/BUFFERSIZE)
        # Second channel: Modulated square wave
        data[1,BUFFERSIZE-x*2:] =  1
        data[1,:BUFFERSIZE-x*2] = -1

        # Write it down
        w.write(data)

Block playback example (using pyaudio)

import pyaudio, sys
from wavefile import WaveReader

p = pyaudio.PyAudio()
with WaveReader(sys.argv[1]) as r:

    # Print info
    print("Title: {r.metadata.title}")
    print("Artist: {r.metadata.artist}")
    print(f"Channels: {r.channels}")
    print(f"Format: 0x{r.format:x}")
    print(f"Sample Rate: {r.samplerate}")

    # open pyaudio stream
    stream = p.open(
        format = pyaudio.paFloat32,
        channels = r.channels,
        rate = r.samplerate,
        frames_per_buffer = 512,
        output = True,
    )

    # iterator interface (reuses one array)
    # beware of the frame size, not always 512, but 512 at least
    for frame in r.read_iter(size=512):
        stream.write(frame.flatten(), frame.shape[1])
        sys.stdout.write("."); sys.stdout.flush()

    stream.close()

Block processing example

import sys
from wavefile import WaveReader, WaveWriter

with WaveReader(sys.argv[1]) as r:
    with WaveWriter(
        'output.wav',
        channels=r.channels,
        samplerate=r.samplerate,
    ) as w:
        w.metadata.title = r.metadata.title + " II"
        w.metadata.artist = r.metadata.artist

        for data in r.read_iter(size=512):
            sys.stdout.write("."); sys.stdout.flush()
            w.write(.8*data)

read_iter simplifies the code by transparently:

  • allocating the data block for you,
  • reusing it for each block, and
  • slicing it when the last incomplete block arrives.

Arquitecture

The library consists of two layers

  • libsndfile.py: a plain ctypes based wrapper that provides 1:1 access to the functions of the sndlib library
  • wavefile.py: an interface layer that provides the Pythonic sugar calling the former one.

Existing alternatives (what i like and dislike)

This is 'yet another' wrapper for sndfile. A lot of them appeared just because the standard 'wave' module is quite limited on what and how it does. But none of the wrappers I found around fully suit my needs and that's because I wrote this small and incomplete one, to fit my needs. So this is a summary of what I found, just in case it is useful to anyone.

python-wavefile reuses most of the libsndfilectypes ctypes wrapper, as not requiring module compilation was seen as a good point. A pythonic layer was added on the top of it.