FFmpeg (contrib)

[TOC]

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