[FFmpeg-user] ffmpeg segmenter | Unable to map audio language metadata

Moritz Barsnick barsnick at gmx.net
Fri May 29 13:44:01 CEST 2015


Hi Tarun,

On Fri, May 29, 2015 at 16:34:08 +0530, tarun singhal wrote:
> ffmpeg version N-61286-gdbc3e11 Copyright (c) 2000-2014 the FFmpeg developers
>   built on Mar 11 2014 22:06:56 with gcc 4.8.2 (GCC)

This version you are using is very old by ffmpeg standards. Please pick
up a recent binary, or build latest git HEAD, and try again before
reporting back. (ffmpeg is under heavy development. Any shortcomings
may have been fixed in the meantime.)

Please don't just look at ffmpeg's output during encoding, but also
check the resulting files with ffmpeg/ffprobe. I can confirm your
observation though.

> -map_metadata:s:a 0:g

I don't know much about "-map_metadata", but I think with 0:g you are
referring to the global metadata, while the language is obviously
specified in the stream metadata. You should probably be using "0:s:a".
Actually, I just tested with that, and it "works for me (TM)". :-)

(If you have multiple audio input streams, you need to use
something like "0:s:a:2" or "0:s:3", otherwise it seems to choose the
metadata of the first audio stream.)

Moritz


More information about the ffmpeg-user mailing list