Ignore audio stream if it is not found in filter_complex of ffmpeg

Question

I am creating a video output using some overlays and scale layers through filter_complex of ffmpeg. I am using ffmpeg v3.3.2

I am using following command for videos with audio:

# avideo -> aoutput : 50% Faster
ffmpeg3 \
    -i avideo.wmv \
    -filter_complex "[0:v]setpts=0.5*PTS[v];[0:a]atempo=2.0[a]" \
    -map "[v]" \
    -map "[a]" \
    aoutput.wmv

I am using following command for videos without audio:

# video -> output : 50% Faster
ffmpeg3 \
    -i video.wmv \
    -f lavfi -t 0.1 -i anullsrc \
    -filter_complex "[0:v]setpts=0.5*PTS[v]" \
    -map "[v]" \
    output.wmv

This works fine. But, I don't know if a video file has any audio streams in it or not. I need to make this command work for both of these scenarios.

I am trying to use following command for this:

# unknown(avideo, video) -> uoutput : 50% Faster
ffmpeg3 \
    -i video.wmv \
    -f lavfi -t 0.1 -i anullsrc \
    -filter_complex "[0:v]setpts=0.5*PTS[v];[0:a]atempo=2.0[a]" \
    -map "[v]" \
    -map "[a]" \
    uoutput.wmv

But, this fails because input video doesn't have any audio stream. Exact error message is

Stream specifier ':a' in filtergraph description [0:v]setpts=0.5*PTS[v];[0:a]atempo=2.0[a] matches no streams.

This question is most similar to my question: FFmpeg map optional (audio) stream, except that they aren't using filter_complex to execute this optional scenario. So, conditional mapping option wouldn't work for me.

How can I tell ffmpeg to use audio stream, in filter_complex, if there's any audio, otherwise ignore audio.

• I want to do this in a single command.
• I want to use ffmpeg alone (no ffprobe: size issues).

But, if you have any good options, please share them.

Edit:

I have updated my command with "-af atempo=2.0" and this seems to fix my issue. I do it like this.

ffmpeg3 \
    -i video.wmv \
    -af atempo=2.0 \
    -filter_complex "[0:v]setpts=0.5*PTS" \
    uoutput.wmv

But, I am still open to more options. Above fix seems wrong. While this works for all formats, but fails for WMV of 30sec with a bug https://trac.ffmpeg.org/ticket/6375

Answer 1

There are two possible hacks for this.

1) If you don't combine processed streams inside the filtergraph, then just use simple filterchains.

ffmpeg \
-i avideo.wmv \
-vf "setpts=0.5*PTS" \
-af "atempo=2.0" \
aoutput.wmv

The audio chain will only be parsed if an an audio output stream is present.

2) Add a dummy audio input and leave out the input label for the audio filtering. ffmpeg will pick the first available audio stream.

ffmpeg \
    -i video.wmv \
    -f lavfi -t 0.1 -i anullsrc \
    -filter_complex "[0:v]setpts=0.5*PTS[v];atempo=2.0[a]" \
    -map "[v]" \
    -map "[a]" \
    uoutput.wmv

I recommend you upgrade to ffmpeg 4.0 or later. There were some filtering bugs in 3.x series, since fixed.