I download old rare European movies sometimes without subtitles.

Some separate subtitle *.srt files can be downloaded in some distant
language for translation into English via Google Translate.

I can't get it to work. Has anyone done this successfully?

Peter Jason <pj@jostle.com> wrote:
> I download old rare European movies sometimes without subtitles.
>
> Some separate subtitle *.srt files can be downloaded in some distant
> language for translation into English via Google Translate.
>
> I can't get it to work. Has anyone done this successfully?

I tried this a few years ago for a German documentary which only
had German subtitles.

While Google Translate works on the raw text (I can't remember
whether the SRT format was used, but I think there was some minor
issue with it changing/adding a character in the time code
formatting which had to be fixed with a find-and-replace-text
operation afterwards), it limits the length of text that can be
converted at once.

There's a website that allows you to submit longer text to Google's
web page translator, which doesn't limit the length, however the
quality of that translation turned out to be must worse. It was
basically gibberish. The lack of context probably makes subtitles
more challenging than a regular web page to translate.

So I broke up the text into individual sections within the allowed
length and manually copied them in/out of the Google Translate
page.

Though much better than the web page translator, the result still
proved quite tiring to watch because of the thought involved in
making sense of the translation. But the documentary was all
interviews, and covered technical topics, so a movie would have a
lot more visual cues and probably more easily translated dialogue.
I'd say it's worth a try.

I don't know if I tried Yandex Translate, but it allows twice as
much text to be processed at once:
https://translate.yandex.com/

There's also Google Cloud Tanslation which has _much_ higher
limits, but it's a paid service aimed at web developers. You can
get free trial access by making an account though.
https://cloud.google.com/translate/

And Yandex have their equivalent to that as well:
https://cloud.yandex.com/en/docs/translate/

--
__ __
#_ < |\| |< _#

Computer Nerd Kev <not@telling.you.invalid> wrote:
> Peter Jason <pj@jostle.com> wrote:
>> I download old rare European movies sometimes without subtitles.
>>
>> Some separate subtitle *.srt files can be downloaded in some distant
>> language for translation into English via Google Translate.
>>
>> I can't get it to work. Has anyone done this successfully?
>
> I tried this a few years ago for a German documentary which only
> had German subtitles.

I found my notes on this (from 2019). I was using SRT format
(although converted from TTML, as pulled out of YouTube using
youtube-dl). I didn't copy into the Text translator box as it
turns out.

The Google document translator had higher length limits, but using
Firefox you still won't do more than the first 30KB of text at a
time so I still had to split it into multiple files (Chrome
apparantly didn't have this limit, so most people would probably
just use it instead).

You have to rename the file extension from ".srt" to ".txt" for
Google to accept it.

The bit I needed to fix with the "find and replace text" operation
was that Google Translate replaced the "-->" bits with "->", which
breaks the decoding.

--
__ __
#_ < |\| |< _#

On Fri, 1 Oct 2021 23:18:46 -0000 (UTC), not@telling.you.invalid
(Computer Nerd Kev) wrote:

>Peter Jason <pj@jostle.com> wrote:
>> I download old rare European movies sometimes without subtitles.
>>
>> Some separate subtitle *.srt files can be downloaded in some distant
>> language for translation into English via Google Translate.
>>
>> I can't get it to work. Has anyone done this successfully?
>
>I tried this a few years ago for a German documentary which only
>had German subtitles.
>
>While Google Translate works on the raw text (I can't remember
>whether the SRT format was used, but I think there was some minor
>issue with it changing/adding a character in the time code
>formatting which had to be fixed with a find-and-replace-text
>operation afterwards), it limits the length of text that can be
>converted at once.
>
>There's a website that allows you to submit longer text to Google's
>web page translator, which doesn't limit the length, however the
>quality of that translation turned out to be must worse. It was
>basically gibberish. The lack of context probably makes subtitles
>more challenging than a regular web page to translate.
>
>So I broke up the text into individual sections within the allowed
>length and manually copied them in/out of the Google Translate
>page.
>
>Though much better than the web page translator, the result still
>proved quite tiring to watch because of the thought involved in
>making sense of the translation. But the documentary was all
>interviews, and covered technical topics, so a movie would have a
>lot more visual cues and probably more easily translated dialogue.
>I'd say it's worth a try.
>
>I don't know if I tried Yandex Translate, but it allows twice as
>much text to be processed at once:
>https://translate.yandex.com/
>
>There's also Google Cloud Tanslation which has _much_ higher
>limits, but it's a paid service aimed at web developers. You can
>get free trial access by making an account though.
>https://cloud.google.com/translate/
>
>And Yandex have their equivalent to that as well:
>https://cloud.yandex.com/en/docs/translate/

Thank you. I get old movies from eMule and the like, preferable the
Criterion prints. I know Google can copy a whole large translation
from a *.srt file (one has to log in) and so far a complicated such
file with color & size coding for the text in the file is ignored by
that translator. Still, I cannot get it to work and the closest
result is just a time code for the first frame. I will try Yandex
next.
I want to try this too....
https://www.youtube.com/results?search_query=translate+srt+file+

Peter Jason <pj@jostle.com> wrote:
>
> Thank you. I get old movies from eMule and the like, preferable the
> Criterion prints. I know Google can copy a whole large translation
> from a *.srt file (one has to log in) and so far a complicated such
> file with color & size coding for the text in the file is ignored by
> that translator. Still, I cannot get it to work and the closest
> result is just a time code for the first frame.

Check it in a text editor to see whether it's a problem like I had
with the "-->" bits replaced by "->". Then you can easily fix it in
a text editor that has a "bulk find an replace text" function.

I refuse to tell Google all the info they want to create an
account, so renaming the files to .txt is the only way I could
upload them.

--
__ __
#_ < |\| |< _#