Extract all subtitles from a movie using ffprobe & ffmpeg

.bashrc

alias subs=subs

function subs() {
    movie="${1}"
    filename="${1%.*}"
    mappings=`ffprobe -loglevel error -select_streams s -show_entries stream=index:stream_tags=language -of csv=p=0 "${movie}"`
    OLDIFS=$IFS
    IFS=,
    ( while read idx lang
    do
        echo "Exctracting ${lang} subtitle #${idx} from ${movie}"
        ffmpeg -nostdin -hide_banner -loglevel quiet -i "${movie}" -map 0:"$idx" "${filename}_${lang}_${idx}.srt"
    done <<< "${mappings}" )
    IFS=$OLDIFS
}

Hi @anhthoai
To debug, this issue, I'd suggest to start with printing out the mappings before the while loop.
Simply, type: echo ${mappings} before the OLDIFS=$IFS line, and re-run the script.
If you're going to see something like: 2,eng 3,fre 4,rus it means that the video file has the language tags correctly set and the index (idx) to language mappings are properly defined.

If you're going to see something different, then please post your output.

The other thing, that I'd check is the format of the embedded subtitles.
You can display it with ffprobe:

ffprobe -loglevel error -select_streams s -show_entries stream=codec_name,index:stream_tags=language -of csv=p=0 "your-video.mkv"
2,subrip,eng
3,subrip,fre
4,subrip,rus

If you're going to see the subtitle format other than subrip then you might need to change the extension of the extracted subtitle from ${idx}.srt to:

• ${idx}.ass for ass (also know as SubStationAlpha)
• ${idx}.pgs for hdmv_pgs_subtitle
• etc

If an input file has more than one embedded subtitle, you can save a lot of time by building a command line that makes a single invocation of ffmpeg extract all of them in one pass.

function subs() {
  local movie idx lang subs=
  local -a dests=()
  for movie
  do
    ffprobe -loglevel error -select_streams s -show_entries stream=index:stream_tags=language -of csv=p=0 "$movie" |
    {
      while IFS=, read idx lang
      do
        subs+=" ${lang}_$idx"
        dests+=(-map "0:$idx" "${movie%.*}_${lang}_$idx.srt")
      done
      if test -n "$subs"
      then
        echo "Extracting subtitles from $movie:$subs"
        ffmpeg -nostdin -y -hide_banner -loglevel quiet -i "$movie" "${dests[@]}"
      else
        echo "No subtitles in $movie"
      fi
    }
  done
}

Rather than have the function process only its first argument, there's an outer for loop that walks through all of them so you can pass multiple filenames explicitly and/or use wildcards.

The list of available subtitle streams is piped straight from the output of ffprobe into the while read loop that parses them, rather than being collected into a shell variable in between. This requires that the subs and dests[] variables that the loop fills in are then used within the same brace-delimited compound statement as the loop itself: the shell runs pipeline components in parallel, each in its own subshell. If it were only the actual while loop that had the ffprobe output piped into it, like

ffprobe -loglevel error -select_streams s -show_entries stream=index:stream_tags=language -of csv=p=0 "$movie" |
while IFS=, read idx lang
do
  subs+=" ${lang}_$idx"
  dests+=(-map "0:$idx" "${movie%.*}_${lang}_$idx.srt")
done

then the loop's subshell would terminate with the loop, and the changes made to subs and dests[] inside the loop would be lost. Note also the use of environment variable prefix syntax to supply a modified IFS to a single command only (the read builtin, in this instance), making it clearer why IFS is being modified and removing the need for an explicit save and restore.

Building dests[] as an array with each of what will become ffmpeg arguments as an individual element, rather than simply appending pieces to a single string as is done with subs, allows those arguments to be cleanly expanded into the ffmpeg command line as separate words even if they contain whitespace or special shell characters. Trying to do this kind of thing in a strictly POSIX-compatible shell that doesn't support arrays is always a nightmare of horrible edge cases. Just don't.

The issue im having is this producing 0kb sub files.

#!/usr/bin/env bash

function subs() {
  local movie idx lang subs=
  local -a dests=()
  for movie
  do
    ffprobe -loglevel error -select_streams s -show_entries stream=index:stream_tags=language -of csv=p=0 "$movie" |
    {
      while IFS=, read idx lang
      do
        subs+=" ${lang}_$idx"
        dests+=(-map "0:$idx" "${movie%.*}_${lang}_$idx.srt")
      done
      if test -n "$subs"
      then
        echo "Extracting subtitles from $movie:$subs"
        ffmpeg -nostdin -y -hide_banner -loglevel quiet -i "$movie" "${dests[@]}"
      else
        echo "No subtitles in $movie"
      fi
    }
  done
}

subs "${1}"