Total length of all audio files (various formats) in directory (with several layers of subfolders) (Linux)

Question

I have a large directory with various kinds of audio files, mp3, ogg, opus, m4a, and possibly others.

I'd like to get the combined length of all these files. If it were only a small number of files, I'd copy them all into another folder and use mp3info, but it's several hundred GB so that's not an option.

I don't think I can use mp3info in my usecase as is, because I can't find an option for it to search recursively. mp3info2 seems to have a recursive option with -R but I can't find a lot of info on mp3info2 so I'm not sure how to use it.

I've tried this

tot=0; while read -r i; do tmp=0;  tmp=`ffprobe "$i" -show_format 2>/dev/null | grep "^duration" | cut -d '=' -f 2 | cut -d '.' -f 1`; if [ -n "$tmp" ]; then let tot+=$tmp; fi;    done < <(find . -type f -iname "*[.mp3,.wav,.m3u,.m4a,.m4b,.mpga,.opus,.opus]"); echo "Total duration: $(($tot/60)) minutes"

but get

bash: let: tot+=N/A: division by 0 (error token is "A")
bash: let: tot+=N/A: division by 0 (error token is "A")
bash: let: tot+=N/A: division by 0 (error token is "A")
bash: let: tot+=N/A: division by 0 (error token is "A")

repeating.

I've tried soxi -D *.mp3 and then would do the different file types but get this

soxi FAIL formats: can't open input file `*.mp3': No such file or directory

The directory format is Letter/Author/Book Title, e.g. K/King, Stephen/The Stand/The Stand.mp3

As a bonus question: How would I do this same thing for video files (in a different directory)

Thanks

Answer 1

It's most probably because your find will find a lot of files that are not multimedia files. When encountering such a file, ffprobe may output something like this:

[FORMAT]
filename=./pong.cpp
nb_streams=1
nb_programs=0
format_name=xbm_pipe
format_long_name=piped xbm sequence
start_time=N/A
duration=N/A
size=17182
bit_rate=N/A
probe_score=99
[/FORMAT]

As you can see, duration is here N/A.

• First, fix the find command. Either use multiple -inames with -o (or) between, or use -regex instead of iname if your find supports it.
• Even if you've found a file with the correct filename, it may be corrupt and the duration may still come back as N/A so make the check safer by checking that you got a number back.

Example:

#!/bin/bash

tot=0
while read -r i; do
    tmp=$(ffprobe "$i" -show_format -loglevel -8 |
          sed -nE 's/^duration=([0-9]+).*$/\1/p')

    if [[ $tmp =~ ^[0-9]+$ ]]; then
        (( tot+=tmp ))
    fi
done < <(find . -type f -iname '*.mp3' -o -iname '*.wav' -o -iname '*.m3u' -o \
                        -iname '*.m4[ab]' -o -iname '*.mpga' -o -iname '*.opus')

echo "Total duration: $((tot/60)) minutes"

---

An alternative could be to collect all the durations, including the subsecond part, in an array and then form an arithmetic expression that you pass on to bc:

#!/bin/bash

join() {
    local IFS="$1"
    shift
    echo "$*"
}

readarray -t durations < <(
    find . -type f \( -iname '*.mp3' -o -iname '*.wav' -o -iname '*.m3u' -o \
                      -iname '*.m4[ab]' -o -iname '*.mpga' -o -iname '*.opus' \) \
                      -exec ffprobe {} -show_format -loglevel -8 \; |
    sed -nE 's/^duration=([0-9\.]+)$/\1/p')

# add + between all durations:
expression=$(join + ${durations[@]})

# calculate the total, including subseconds, round to whole minutes:
tot=$(bc -q <<< "scale=0;($expression)/60")

echo "Total duration: $tot minutes"