Recommendations for bilingual SRT editors

[Doitsujin]

Does anybody have recommendations for SRT editors that allow users to align two SRT files? I'm in particular interested in editors that allow users to copy the translations over the original text so that both have the same time codes.

So far I've only found one tool, Subs Re-Timer, which comes with sub2srs. However, it doesn't allow users to merge or split source or target segments. (Merging is necessary, if the source SRT file contains one segment and the target SRT file two or more segments; splitting is necessary in the opposite case.)

Here's an example of two German segments that would need to be merged:



Are there any other free tools like Subs Re-Timer that support merging/splitting of source and target segments?

[rdearman]

We have a thread about this a long while ago. I did manage to align SRT files (with help from EMK's instructions) using a free linux tool called Aeneas. I didn't bother with the VM Docker thing, but just installed it directly on my Linux box. You can also look at hunalign.

https://forum.language-learners.org/vie ... =14&t=7314

[Adrianslont]

Have you looked at the free program SubtitleEdit. It’s for windows.

I never align two languages, I just use it’s built in Google translate function and get a very literal translation BUT as far as I can see, you can align.

It will split and merge a line, easily.

[白田龍]

I wrote this in Python it should copy the timings and merge... but it will make a mess when split would be required...

needs: https://github.com/tkarabela/pysubs2

Code: Select all

import pysubs2


engSub = pysubs2.load(r'timingSub.ass')
japSub = pysubs2.load(r'textSub.ass')

lastTrans = (0.0, 0.0)
def getTranslation(line, sub):
   global lastTrans
   return u'\\N'.join([a.text for a in sub if (a.start + a.end)/2 >= line.start and \
   (a.start + a.end)/2 <= line.end and \
   (a.start,  a.end) != lastTrans])
   
   lastTrans = (a.start,  a.end)



for line in engSub:
   line.text = getTranslation(line, japSub)

engSub.save("result.srt")

[Kraut]

https://github.com/bonigarcia/dualsub/

i don't know if Google Translate now works for Dualsub, but the two subs should be aligned matchingly.

[iguanamon]

For those of us who do not have programming skills, there's another way to align subs. It can be done manually. Once a learner gets to the point where they can benefit from subs, they can be downloaded into an srt file and opened in microsoft word pad.

Then open a word/openoffice document. Insert a two column one row table. Insert TL subs on the left. Then insert L1 subs on the right and line them up. The time stamps almost never match. sometimes there may be extraneous data- like written signs being translated that are not in the original audio; stage comments: "she coughs"; etc. These superfluous phrases can be easily omitted manually.

It's funny, many learners here rave about how useful Assimil is. An Assimil course is a parallel text and annotated graded reading (with audio). While it isn't easy to replicate the graded reading from online sources or the annotations, the parallel text aspect is easy to replicate, even if one doesn't have the computer skills necessary to do it. I believe the reason more learners do not take advantage of parallel texts is that they are not easily available out of the box, "plug and chug" ready.

Even if a learner has little knowledge of TL, clues to help with manual alignment are present with proper names, international vocabulary, etc. I can align a 22 minute dubbed US show in about 5-10 minutes work in a Latin script IE language. I can print it to pdf, add notes and definitions, put it on my phone or tablet and read it while I listen on another device... if I have audio. If my pronunciation is good enough, I don'e even need the audio.

While I know that it is much easier and desirable to let the computer do the work for you, and I know we have a large number of members who know how to program, there are also a lot of us who simply do not have these skills and wouldn't know a Python script from Devanagari script. Learning from parallel texts, whether they be from subtitle files or short articles, is so highly useful and very often underutilized by many learners. It is a wasted opportunity for many learners who could gain a lot of benefit from using them in their learning. Integrating parallel texts in learning languages can add a major boost in learning a language, bridging the gap in reading when a learner doesn't yet know a large percentage of vocabulary to make reading less frustrating.

I wouldn't do a manual alignment of a book, that would take a lot of time to do manually, but a half an hour or an hour television show, a short 3-5 page article, a book of the Bible, yes. That's easy to do, even if I can't program my way out of a wet paper sack.

[rdearman]

Good point. I did actually align an entire book once. I didn't use word but a linux editor but it worked out the same. I didn't use tables but page breaks so that when printed I had english on odd pages and french on even pages.

EDIT: I thought I would show some photos. Admittedly the OP wanted electronic files, but you can generate dead tree materials too. It took me about 3-4 hours if I remember correctly, but afterwards it was relativity easy to produce a PDF, print, glue, and trim into a nice book. Which reminds me, I should dig it out and read it. I actually made about 4 of them and gave 3 away to friends who were learning French.

thumb_book1.jpg

thumb_book2.jpg

[Kraut]

There is a problem with professionally done translations that can be seen in the first line of the example above

He told me of countless insults suffered at the hands of

Er erzählte von Beleidigungen seitens sein ....

Three words don't appear in the German translation at all: "me", "countless" and "suffered"; "at the hands of" is reduced to the single word "seitens".
Comparing this may be interesting to a student of stylistics, but not to a learner.
I also do the bi-directional method, but always look for an equivalent as long as it doesn't hurt grammar (and style) too much.
It can also be done here:

Er erzählte mir von zahllosen Beleidigungen , die er von Seiten ... erlitten hatte.

[Doitsujin]

Three words don't appear in the German translation at all: "me", "countless" and "suffered"; "at the hands of" is reduced to the single word "seitens".

Subtitlers need to observe line llimits specified by movie production companies and/or distributors. For example, Netflix recommends around 42 Latin characters per line. And since German words tend to be longer than English words, subtitlers are often forced to omit some words in order not to exceed these limits.

[mcthulhu]

There are a number of sentence alignment tools available, used for producing bilingual translation memories (databases of aligned sentences), for instance, which allow splitting and merging segments. This task seems similar enough. LF Aligner (free) at https://sourceforge.net/projects/aligner/ would probably be my first recommendation. It has a graphical interface for correcting alignments. For a commercial tool, see https://youalign.com/.

Once you have a .tmx file, a TMX editor like https://sourceforge.net/projects/tmxeditor/ might also be useful for correcting it.

Obviously you can do anything you want manually, if you have enough time, but it's a lot more work than using a tool designed for that purpose.