Convert between audio formats

Convert in CLI

sox new.wav new.mp3

Convert to a different bit depth and sampling rate

ffmpeg -i  input.wav -ac 1 -ar 16000 -sample_fmt s16 output.wav

Available bit depth optons:

ffmpeg -sample_fmts

Convert in Python

import subprocess

in_file = "test.webm"
out_file = "test.wav"

subprocess.run(
    [
        "ffmpeg",
        "-i",
        in_file,
        "-ar",
        "16000",
        "-ac",
        "1",
        "-sample_fmt",
        "s16",
        out_file,
    ],
    stdout=subprocess.PIPE,
    stderr=subprocess.STDOUT,
)

Convert Numpy to WAV with wavio

wavio.write writes a numpy array to a WAV file, optionally using a specified sample width.

Create a numpy array of 16-bit sine wave

import numpy as np

frequency = 440  # Our played note will be 440 Hz
fs = 44100  # 44100 samples per second
seconds = 3  # Note duration of 3 seconds
t = np.linspace(0, seconds, seconds * fs, False)
note = np.sin(frequency * t * 2 * np.pi)
audio = note * (2**15 - 1) / np.max(np.abs(note)) # Ensure that highest value is in 16-bit range
my_np_array = audio.astype(np.int16) # Convert to 16-bit data

Save to wav file

import wavio
wavio.write('/tmp/myfile.wav', my_np_array, fs, sampwidth=2)

!ls /tmp/myfile*.wav

/tmp/myfile.wav

Convert with pydub

Can convert between all ffmpeg formats

from pydub import AudioSegment
sound = AudioSegment.from_file('/tmp/myfile.wav', format='wav')
sound.export('/tmp/myfile.mp3', format='mp3')

<_io.BufferedRandom name='/tmp/myfile.mp3'>

!ls /tmp/myfile*

/tmp/myfile.mp3 /tmp/myfile.wav

Convert with soundfile

Can convert between all libsndfile formats

import soundfile as sf

# Extract audio data and sampling rate from file 
data, fs = sf.read('/tmp/myfile.wav') 
# Save as FLAC file at correct sampling rate
sf.write('/tmp/myfile.flac', data, fs)  

!ls /tmp/myfile*

/tmp/myfile.flac /tmp/myfile.mp3  /tmp/myfile.wav