Automatically extract ALL audio streams from a video file into separate files?

Question

I have a .mov video file, for which ffprobe/avprobe reports:

$ avprobe -show_streams test.mov

    Stream #0.0(eng): Video: dvvideo, yuv420p, 720x576 [PAR 64:45 DAR 16:9], 28811 kb/s, PAR 236:162 DAR 295:162, 25.01 fps, 2500 tbn, 25 tbc
    Metadata:
      creation_time   : 2015-08-27 09:10:44
    Stream #0.1(eng): Audio: pcm_s16le, 48000 Hz, 1 channels, s16, 768 kb/s
    Metadata:
      creation_time   : 2015-08-27 09:10:44
    Stream #0.2(eng): Audio: pcm_s16le, 48000 Hz, 1 channels, s16, 768 kb/s
    Metadata:
      creation_time   : 2015-08-27 09:10:44
    Stream #0.3(eng): Data: tmcd / 0x64636D74
Unsupported codec with id 0 for input stream 3
...
[streams.stream.1]
index=1
codec_name=pcm_s16le
codec_long_name=PCM signed 16-bit little-endian
codec_type=audio
codec_time_base=1/48000
codec_tag_string=sowt
codec_tag=0x74776f73
sample_rate=48000.000000
channels=1
bits_per_sample=16
...
[streams.stream.2]
index=2
codec_name=pcm_s16le
codec_long_name=PCM signed 16-bit little-endian
codec_type=audio
codec_time_base=1/48000
codec_tag_string=sowt
codec_tag=0x74776f73
sample_rate=48000.000000
channels=1
bits_per_sample=16
...
[streams.stream.3]
index=3
codec_name=unknown
codec_type=data
codec_time_base=0/1
codec_tag_string=tmcd
codec_tag=0x64636d74
avg_frame_rate=0/0
time_base=1/2500
start_time=0.000000
duration=32.760000
nb_frames=1
...

Now, from this I cannot tell how the audio is encoded (if it isn't uncompressed), and I cannot tell whether these three audio streams are some sort of a surround sound (and one of them cannot be detected either) - but I guess they are because mediainfo reports for the same file:

Audio #1 ...
Codec ID                                 : sowt
Channel(s)                               : 1 channel
Channel positions                        : Front: C
Sampling rate                            : 48.0 KHz
Bit depth                                : 16 bits
...
Audio #2 ...
Codec ID                                 : sowt
Channel(s)                               : 1 channel
Channel positions                        : Front: C
Sampling rate                            : 48.0 KHz
Bit depth                                : 16 bits
...
Audio #3 ...
Muxing mode, more info                   : Muxed in Video #1
Channel(s)                               : 2 channels
Sampling rate                            : 48.0 KHz
Bit depth                                : 16 bits
...

So, is there a Linux command line that I could use, to extract all of these audio streams from the video file in one go; that is, this pseudocode command:

my_command --extract-all-audio-streams test.mov

... should result, in this case, with three audio files corresponding to the three streams.

In particular:

• What command could I use, so that the streams are extracted unconverted from the file, akin to -acodec copy of ffmpeg - and so that file extensions are automatically assigned (i.e. if it is uncompressed audio, then .wav, if it is AAC encoded, then .aac etc)
• What command could I use, so that the streams are extracted converted as uncompressed audio, of the same settings as the streams? That is, the resulting test_s01.wav would be mono/48.0 KHz/16 bits, test_s02.wav would be also mono/48.0 KHz/16 bits, and test_s03.wav would be stereo/48.0 KHz/16 bits?

Answer 1

Well found something with ffmpeg/avconv and .wav files; it's not exactly automatic, but it is a one-liner extracting the audio. First, use ffprobe/avprobe as in OP, or ffmpeg -i/avconv -i to get the stream info; from the OP:

Stream #0:0(eng): Video:
Stream #0:1(eng): Audio:
Stream #0:2(eng): Audio:
Stream #0:3(eng): Data:

So, the Audio streams are 0:1 and 0:2 (note ffprobe/avprobe writes dot . as separator, ffmpeg -i/avconv -i writes a colon :), and those we want saved as .wav files - so this is the command:

avconv -i test.mov \
  -map '0:1' -vn -acodec copy test-01.wav \
  -map '0:2' -vn -acodec copy test-02.wav

... and at least in this case of mine, the resulting wavs have identical properties with the streams:

$ mediainfo *.wav
...
Channel(s)                               : 1 channel
Sampling rate                            : 48.0 KHz
Bit depth                                : 16 bits
...
Channel(s)                               : 1 channel
Sampling rate                            : 48.0 KHz
Bit depth                                : 16 bits

It now looks to me like these two "mono" "center" channels are actually left and right channels - and the stream 3, even if it's called "Audio", it's still type "data", and so maybe it is just a descriptor of how these L and R mono channels are muxed (as per the mediainfo report "Muxing mode, more info")?!

Still, it would be great to know if there is a tool that could do this sort of stuff automatically...

---

_{I would have thought that one has to use filters, so I've tried this:}

avconv -i test.mov -filter_complex '[0:1] copy [a1]; [0:2] copy [a2]' \
  -map '[a1]' -acodec copy test-01.wav \
  -map '[a2]' -acodec copy test-02.wav

_{... but this fails with Stream specifier ':1' in filtergraph description [0:1] copy [a1]; [0:2] copy [a2] matches no streams., don't really understand why.}