FFmpeg (contrib)

Encoding and decoding audio using FFmpeg

TensorFlow provides Ops to decode and encode audio files using the FFmpeg library. FFmpeg must be locally installed for these Ops to succeed.

Example:

from tensorflow.contrib import ffmpeg

audio_binary = tf.read_file('song.mp3')
waveform = ffmpeg.decode_audio(
    audio_binary, file_format='mp3', samples_per_second=44100, channel_count=2)
uncompressed_binary = ffmpeg.encode_audio(
    waveform, file_format='wav', samples_per_second=44100)

• tf.contrib.ffmpeg.decode_audio
• tf.contrib.ffmpeg.encode_audio