Translation of video subtitles of promo team

[Johnny Jazeix]

Le mer. 2 août 2023 à 19:44, Karl Ove Hufthammer <karl at huftis.org> a écrit :
>
> Johnny Jazeix skreiv 02.08.2023 17:38:
> > I've been discussing with promo on how to help them translate the
> > subtitles in their videos (for example
> > https://tube.kockatoo.org/w/symxHCzUj9TA7LZoZagD9c).
> > In order to not change the process of translators, I told them I can
> > create a script that will convert the srt files (subtitle file) to po
> > files. […]
>
> In case you’re not aware, the Translate Toolkit has a tool for this:
> https://docs.translatehouse.org/projects/translate-toolkit/en/latest/commands/sub2po.html
>

I wasn't aware, thank you! After some tests, it allows to do what we
want except for changing the timestamps.
They are available in the po files created with sub2po, so it could
still be possible to create a script that will do the conversion from
po to srt only if we need to be able to change them.

>
> > I have a few questions regarding the translation process:
> > * should there be only one po file containing all the translations for
> > all subtitles files, or one po file per subtitle file?
>
> That depends. If different videos contain many of the same strings, I
> think it’s better to lump all of the strings into one file (but sort the
> strings by video + timepoint). This will reduce the translators’ work
> and ensure consistent translations across videos. But if the different
> videos contain mostly unique strings, I would prefer a separate file for
> each video. Then it’s easier to choose which videos to translate.
> Translating one giant file easily feels like unsurmountable work …
>

I think it will be mostly different strings. Luigi also told me in the
matrix channel he would prefer one file per video so I'll go with it.
sub2po tool also do a 1 to 1 conversion, so it will avoid merging and splitting.

I think Paul answered all the other points.

Cheers,

Johnny

>
> > * There could be issue with display time of texts. Basically, in the
> > subtitle file, for each line, you tell when the text appears and when
> > it hides. If a translated text takes more time to be read, the user
> > might have to pause the video in order to read it which is a bad
> > experience. To solve this potential issue, I propose to also add in
> > the messages to translate the timestamp of begin and end, so
> > translators can also update them if needed.
>
> Yes, that’s a good idea, at least as long as the video part is not
> changed in the future (so that the subtitles get of of sync). It’s of
> course more work for the translators. But it’s *optional* work; you can
> choose to just use the original timestamps.
>
> Are the videos supposed to have spoken (subtitled) dialogue? Or are the
> subtitles just translations of the English on-screen text?
>
> I have some experience translating subtitles (for Scratch video
> tutorials). It wasn’t possible to change the timestamps. One solution
> was to split the text at a different place in the translation, to limit
> the number of words in each string (and the number of lines to two).
> Then each translated string didn’t correspond exactly to its English
> counterpart (but instead to a neighbouring English string). That worked OK.
>
> Also, one problem I often see, is people mistaking subtitles for
> transcripts. For *natural* speech, real subtitles are *highly condensed*
> versions of the words spoken. If you don’t condense the text, you will
> get subtitles that are too long (to read) and that are shown on the
> screen for just a very short duration. Not very user friendly.
>
> Oh, and one more thing. It must be possible to manually choose the
> placement of the line breaks (using normal ‘\n’ characters).
>
> --
> Karl Ove Hufthammer
>