Extracting Audio channel from a video file to feed in decode_wav function of TensorFlow

Question

I want to feed audio channel of a video file to the following TenorFlow function:

tf.audio.decode_wav(
contents,
desired_channels=-1,
desired_samples=-1,
name=None)

Where Args:

contents: A Tensor of type string. The WAV-encoded audio, usually from a file.
desired_channels: An optional int. Defaults to -1. Number of sample channels wanted.
desired_samples: An optional int. Defaults to -1. Length of audio requested.
name: A name for the operation (optional).

Check those answers: [How does tf.audio.decode_wav get its contents?](https://stackoverflow.com/questions/58096095/how-does-tf-audio-decode-wav-get-its-contents) & [Python extract wav from video file](https://stackoverflow.com/questions/26741116/python-extract-wav-from-video-file). — Xarvalus, Oct 10 '19 at 14:30

score 1 · Accepted Answer · answered Oct 10 '19 at 14:38

You can extract the audio of video by eg.:

import subprocess

command = "ffmpeg -i C:/test.mp4 -ab 160k -ac 2 -ar 44100 -vn audio.wav"

subprocess.call(command, shell=True)

And pass the *.wav file as tensor to tf.audio.decode_wav:

raw_audio = tf.io.read_file(filename)
waveform = tf.audio.decode_wav(raw_audio)

References:

score 0 · Answer 2 · answered Oct 13 '19 at 07:42

For windows 10, I solved my problem this way:

I downloaded and installed ffmpeg for windows from here

Extracted wav audio files from videos as:

import os,shlex, subprocess
files = os.listdir('.')
#convert all video files of the current directory to wav audio files
for f in files:
    if(f.endswith('.wmv') or f.endswith('.mp4') or f.endswith('.avi')):
        command_line = "ffmpeg -i "+f+" -ab 160k -ac 2 -ar 44100 -vn "+f[0:-4]+".wav"
        #command_line = "ffmpeg -i default.wmv -ab 160k -ac 2 -ar 44100 -vn default.wav"
        args = shlex.split(command_line)
        print(args)
        p = subprocess.Popen(args) # Success!
        print(p)

Decoded audio files to wavform as:

    import tensorflow as tf
files = os.listdir('.')
#convert all video files of the current directory to wav audio files
for filename in files:
    if(filename.endswith('.wav')):
        print(filename)
        audio_binary = tf.io.read_file(filename,name=None)
        waveform,_ = tf.audio.decode_wav(
            audio_binary,
            desired_channels=-1,
            desired_samples=-1,
            name=None
        )

Extracting Audio channel from a video file to feed in decode_wav function of TensorFlow

2 Answers2