.github.io

Viewing English and Japanese subtitles at the same time

13th September 2013 · 3 min read

On Linux I don’t know of any good media players that’ll show multiple subtitles together, which poses a problem if you want to watch with subs in both the language you’re learning and your native tongue.

Fortunately the unfortunately-named ASS format supports displaying more than one sub at a time, and we can use that to get a result like this:

EN/JP subs together

The method

In my case, I had a bunch of Japanese subs in .srt format, and English subs in .ass format. The approach is simple: for each video,

  • Create a new .ass layer called ‘Japanese’ alongside default.
  • Change the layer alignments to put one at the top, one at the bottom.
  • Parse the .srt to extract subtitle lines and timings.
  • Generate equivalent lines in .ass format, and shove them on the end of the existing file, using the new ‘Japanese’ layer.

.ass is a great format because subs don’t have to be ordered by time or anything. The only snafu is that there are two formats (‘v4’ and ‘v4+’) which have completely different alignment values (e.g. ‘2’ might correspond to ‘bottom left’ for one format, but ‘top centre’ for another). So we also upgrade any v4 subs to v4+ for sanity reasons.

The goods

Here is an example set of subs for 日本人の知らない日本語 (“Japanese that Japanese people don’t know”):

The script

Here is the script I wrote; it’s pretty specific to one set of subs, but I may expand it once I have tried a few other shows.

#!/usr/bin/perl
use strict;
use warnings;
use autodie ':io';
use utf8;
use open ':std', ':encoding(UTF-8)';
use File::Temp;
use File::Copy qw(move);

my $JP_SPACE = '　';
my $JP_STYLE = 'Style:Japanese,Arial,24,&H00FFEEEE,&H000000FF,&H00000000,&H00000000,0,0,0,0,100,100,0,0,1,2,1,2,10,10,10,1\n';

for my $episode (qw(01 02 03 04 05 06 07 08 09 10 11), '12 finale') {
    my $base = "Nihonjin no Shiranai Nihongo ep$episode (704x396 XviD)";
    convert("$base.ass", "$base.jp.srt", "enjp-$base.ass");
}

sub convert {
    my ($ass, $srt, $out) = @_;
    open my $assfh, '<:utf8', $ass;
    open my $srtfh, '<:utf8', $srt;
    my $outfh = File::Temp->new;
    binmode $outfh, ':utf8';

    # Read .ass file and add Japanese style
    while (<$assfh>) {
        if (/^Style:\s*Default/) {
            # Change alignment from middle-bottom to middle-top and adjust border
            s/2,2,2/2,1,8/;
        }
        # Upgrade to v4+ for sanity reasons - alignment numbers different between v4 and v4+
        s/v4\.00(?!\+)/v4.00+/i if /^ScriptType:/i;
        s/V4/V4+/i if /\[V4 Styles\]/i;
        print {$outfh} $_;
        if (/^Style:\s*Default/) {
            print {$outfh} $JP_STYLE;
        }
    }

    # Now add Japanese lines
    while (<$srtfh>) {
        my $num = $_;
        # A byte-order mark? Get out with yez
        $num =~ s/^\x{FEFF}//;
        $num =~ /^\d+\r?\n$/ or die "can't parse '$_'";
        my $timings = <$srtfh>;
        $timings =~ /^(.*?) --> (.*?)\r?\n$/;
        my ($start, $end) = ($1, $2);
        my @lines;
        while (<$srtfh>) {
            s/\r?\n//g;
            last if !$_;
            push @lines, $_;
        }

        my $all = join '\\n', @lines;
        $all =~ s/$JP_SPACE+/$JP_SPACE/g;

        printf {$outfh} "Dialogue: 0,%s,%s,Japanese,,0000,0000,0000,,%s\n",
            _time2ass($start),
            _time2ass($end),
            $all;
    }

    close $outfh;
    move($outfh->filename, $out);
}

sub _time2ass {
    my $t = shift;
    $t =~ /^(\d+):(\d+):(\d+),(\d\d)\d*$/ or die "invalid time '$t'";
    my ($h,$m,$s,$ms) = ($1, $2, $3, $4);
    $h = 0+$h; # no leading zeros
    return sprintf '%d:%02d:%02d.%02d', $h, $m, $s, $ms;
}

Site note

.sub/.idx files are very common; they’re a huge pain because the subtitles are stored as images and require OCR. The best bet there is probably to hard-code the subs onto the video, then layer the other set over them.