This package provides a simple and unified API to read and write sound-files to and from numpy arrays.
- reads all the samples (or a fragment) from a soundfile and returns a tuplet (data, samplerate)
- Data will always be as a
numpy.float64, between -1 and 1, independently of bit-rate
- reads chunks of frames, avoiding the allocation of all the samples in memory
SndInfo, a namedtuple with all the information of the sound-file
- writes the samples.
- samples need to be a numpy.float64 array with data between -1 and 1
- allows you to write to the file as samples become available
Resample a numpy array to a new samplerate
# Normalize and save as flac from sndfileio import sndread, sndwrite samples, sr = sndread("in.wav") maxvalue = max(samples.max(), -samples.min()) samples *= 1/maxvalue sndwrite(samples, sr, "out.flac")
# Process a file in chunks from sndfileio import * from sndfileio.dsp import with sndwrite_chunked(44100, "out.flac") as writer: for buf in sndread_chunked("in.flac"): # do some processing, like changing the gain buf *= 0.5 writer.write(buf)
pip install sndfileio
apt install libsndfile1-dev)
In windows and macos all external dependencies are installed automatically
See the LICENSE file for license rights and limitations (MIT).