Converting Youtube TimedText (xml/vtt) into SubRip (srt)

I also now understand that it's better to use one single tool to do multiple things at once than calling several different ones.

With pipes between them, not necessarily, because the commands run in parallel. I tried to explain that in the longest paragraph of my previous post.

1. t /= 1000
Does this means "variable t becomes its own value divided by 1000"? I tried running it in a simple terminal window and it gave an error (like this: t=80000; t/=1000; printf "$t").

Yes, t /= 1000 divides t by 1000 and store the result in t. As in C. AWK is essentially a simplified C (no declaration of variables, which are automatically initiallized and transparently converted between strings and float, no pointer, arrays are associative, etc.) mixed with sed (for the structure of the program, the handling of regular expressions and the functions sub and gsub) and with many variables that are automatically defined (NR, FNR, NF, $1 to $NF, $0, etc.). To access the content of the t variable, you just write t, as in C:
$ awk '{ t = 80000; t /= 1000; printf t }'

80
(I typed [Enter] to end a record; Ctrl+D to end the input.)

If you add a dollar sign before the last t, it would print $80, the 80th fields of the records (because t is 80). If you further add double quotes around $t, then it is the string "$t" that is printed (literally "$t", but without the quotes). I have no idea how you ended up with an error.

2. finalfile=$(echo "$1" | rev | cut -c 5- | rev ).srt
(...) how would you suggest we improve on this one?

Well, I would write on the standard input, as I did. In this way you can redirect the output with > (or >> to append) or you can further process it through a pipe.

Nevertheless, to process several subtitle files given as arguments to the script (your last question), you do need to define the output file names. I would wrap the script ("(...)" below, where "$1" must become "$xml") in this way:
for xml in "$@"
do
srt=$(printf "$xml" | sed 's/\.[^.]*$//').srt
(...) > "$srt"
done

sed 's/\.[^.]*$//' deletes everything from the last dot. The shell appends ".srt". I do not substitute with ".srt" in sed to have ".srt" appended even if the file name contains no dot (otherwise the output file would be the input file). Your solution always substitutes the last four characters. It only works for file names with three-letter extensions. I am aware that \.[^.]*$ looks arcane... but to seriously use grep, sed or awk, learning POSIX regular expressions is required.

Notice that performance does not matter here: processing the file names is far less work than processing the files themselves.

"$@" are all the arguments of the shell script. My variable named "xml" successively contains them. Those arguments being file names here, they can contain spaces. That is why I write to "$xml", with the quotes, and not to $xml, without the quotes. In your script, such quotes are missing around the first occurrence of $1.

EDIT (I forgot to explain the problem with the instruction at the end of your post): For the shell script, "$1" is its first argument. If you prefer, you can actually have "$1" successively be every every argument, thanks to shift:
while [ -n "$1" ]
do
srt=$(printf "$1" | sed 's/\.[^.]*$//').srt
(...) > "$srt"
shift
done

Subtitle counting is different

If you want the last AWK program to count the subtitles from 1 instead of copying the numbers from the original subtitles, just replace in it:
print "\n" $1 "\n" $2
with
print "\n" ++nb "\n" $2

Is there any safeguard that could be used in the code to make sure it works with both (if you see any difference)

The difference is that:

• 1.txt was written from an operating system that ends lines with "\n":
  $ head -1 1.txt | hexdump -c
0000000 4 \n
0000002
• t1.txt was written from an operating system (most probably Windows) that ends lines with "\r\n" (hence a waste of one character per line):
  $ head -1 t1.txt | hexdump -c
0000000 1 \r \n
0000003

Below, I added an optional '\r' character before every '\n' in the inputs:
#!/bin/sh
awk -v RS='\r?\n\r?\n' -F '\r?\n' '
{
l += length($3)
len[++i] = ++l }
END {
for (j = 1; j <= i; ++j)
print len[j] / l }' "$1" |
awk -v RS='[ \r\n]+' -v total=$(awk -v RS='[ \r\n]+' '{ l += length } END { print l + NR }' "$2") '
{
if (FILENAME == "-")
proportion[NR] = $0
else {
l += length
if (++l > total * proportion[i]) {
printf "\n\n%s", $0
++i }
else
printf " %s", $0 } }' - "$2" |
tail -n +3 |
awk -v RS='\r?\n\r?\n' -F '\r?\n' '
{
print "\n" $1 "\n" $2
getline < "/dev/stdin"
print }' "$1"
No '\r' remains in the output.

EDIT: counting the total number of characters in the second file with awk (wc counts multibyte characters as several) so that it is certain the total will be reached; in that second file, sequences of ' ', '\r' and '\n' make no difference (they are treated as one single space): who modifies the text can freely use those separators.