wavfile

A lightweight package to read/write wave audio files to/from lists of native Python types.

The package currently supports PCM (integer) and IEEE float formats, and supports arbitrary integer precision, including: 16-, 24-, 32-, and 64-bit samples.

View online documentation at https://pywavfile.readthedocs.io/.

Usage: reading wave files

The package provides wavfile.open() to open audio files for reading and writing. For reading:

f = wavfile.open(file, 'r')

where file is either a path to a wave file or a pointer to an open file. This returns a wavfile.wavread.WavRead object with the following properties:

num_channels: The number of audio channels in the stream.
sample_rate: The sampling rate/frequency of the audio stream.
bits_per_sample: The number of bits per audio sample.
num_frames: The total number of audio frames in the audio stream. A frame is a block of samples, one for each channel, corresponding to a single sampling point.
duration: The duration of the audio file in seconds.
hms: The duration of the audio file formatted as hh:mm:ss.tt.
audio_fmt: The audio sample format.
format: The wave format code.
metadata: A dictionary containing metadata encoded in the file.

The object also has the following methods:

read([N]): Read, at most, N frames from the audio stream in their unmodified native format. The method returns a list of lists with size (N,C), where C is the number of audio channels. Excluding N, choosing N = None or N < 0 will read all remaining samples.
read_int([N]): This method is identical to read() except that it returns the samples as integers using the specified bit depth.
read_float([N]): This method is identical to read() except that it returns the samples as floats in the range [-1, 1).
iter([N]): Similar to read() but used in iterator contexts to read successive groups of audio frames.
iter_int([N]): Similar to read_int() but used in iterator contexts to read successive groups of audio frames.
iter_float([N]): Similar to read_float() but used in iterator contexts to read successive groups of audio frames.
seek(N [, whence]): Move to the Nth frame in the audio stream; whence sets the positioning: 0 (default) = absolute positioning, 1 = relative to current position, 2 = relative to end of last frame.
tell(): Return the current frame in the audio stream.
close(): Close the instance.

Alternatively, the wavfile.read() shortcut function is provided:

audio, sample_rate, bits_per_sample = wavfile.read(file, fmt='int')

where fmt is 'int', 'float', or 'native'; and audio is the audio data. The function reads all audio data in the file.

Usage: writing wave files

f = wavfile.open(file, 'w',
                 sample_rate=44100,
                 num_channels=None,
                 bits_per_sample=16,
                 fmt=wavfile.chunk.WavFormat.PCM)

where sample_rate is the sampling rate for the new file, num_channels is the number of audio channels, bits_per_sample is the number of bits used to encode each sample, and fmt is the wave format code. If num_channels is unspecified it will be determined automatically from the first block of samples that are written (see below). This returns a wavfile.wavwrite.WavWrite object. The object shares its properties with the wavfile.wavread.WavRead class.

The format code is either wavfile.chunk.WavFormat.PCM, wavfile.chunk.WavFormat.IEEE_FLOAT, or wavfile.chunk.WavFormat.EXTENSIBLE. The extensible code corresponds to a variation of the wave file format intended for audio with: a bit depth of greater than 16 bits, or more than two channels. The file will be updated automatically to use the extensible format as appropriate. If the extensible format is specified explicitly, then the audio data will be PCM encoded. The audio_fmt property always reports the audio sample format, whereas the format property reports the format code that may be any of the three aforementioned values.

The object also offers the same seek(), tell(), and close() methods. In addition, the following methods are provided for writing audio data:

write(audio): Write frames of audio data to the audio file. The data should be contained in a list of lists with size (N,C), where N is the number of frames and C is the number of audio channels. The data may be int or float. The data may be converted if they do match the format of the destination file.
write_int(audio): Write frames of integer audio data to the audio file. The data may be converted if they do match the format of the destination file.
write_float(audio): Write frames of float audio data to the audio file. The data may be converted if they do match the format of the destination file.
add_metadata(**kwargs): Add metadata to the .wav file.

Alternatively, the wavfile.write() shortcut function is provided:

wavfile.write(file,
              audio,
              sample_rate=44100,
              bits_per_sample=16,
              fmt=wavfile.chunk.WavFormat.PCM,
              metadata=None)

where audio is the audio data to write to the file.

Installation

From source:

python -m pip install -e --user .

From PyPI:

python -m pip install --user wavfile

License

wavfile is available under the MIT license. See LICENSE.txt for more information.