Post your JAV subtitle files here - JAV Subtitle Repository (JSP)★NOT A SUB REQUEST THREAD★

sadpandajav

New Member
Jun 27, 2022
5
2
Here is the English subtitle for IPX-889 Minami Aizawa


View attachment 2964512


Credits goes to the original subber for Supjav. Tools used:
VSF: colour filters to pick the coloured and moving subs;
SubtitleEdit: sync'd with the proper video;
Naver and DEEPL for translation
Could you explain the settings used in VSF to deal with the coloured and moving subs? I didnt think it was possible with the latest batch.
 

mei2

Well-Known Member
Dec 6, 2018
246
405
ssis00434pl.jpg

My first attempt to transcribe from amazon. it's much better than autosub but still not perfect. I am trying to get rid of my reliance on hardcoded Chinese sources. As usual use it only as contextual and fill in the blanks with your imagination.


Thanks! Will check it out right away.
Do you have the original ja-JP version --would be interesting to compare the translation.
 

alexi395

Member
Jul 31, 2020
38
63
Could you explain the settings used in VSF to deal with the coloured and moving subs? I didnt think it was possible with the latest batch.
Here is the English subtitle for IPX-889 Minami Aizawa


View attachment 2964512


Credits goes to the original subber for Supjav. Tools used:
VSF: colour filters to pick the coloured and moving subs;
SubtitleEdit: sync'd with the proper video;
Naver and DEEPL for translation
VSF: color filters ,I tried many settings but wasn't satisfied with the results ,managed to only get around 100 lines out of 500+. Can you share your conf file from VSF if possible?
Thanks
 

mei2

Well-Known Member
Dec 6, 2018
246
405
VSF: color filters ,I tried many settings but wasn't satisfied with the results ,managed to only get around 100 lines out of 500+. Can you share your conf file from VSF if possible?
Thanks
Here is the config file. I used the following 4 colors:

Lab: l:205 a:133 b:100;
Lab: l:216 a:147 b:133;
Lab: l:230 a:94 b:151;
Lab: l:249 a:113 b:159

I tested it on IPX-889 and it resulted quite well --more than 85% complete+correct detection I think.
 

Attachments

  • VSF_cfg_w_4_colours_supjave.zip
    1.1 KB · Views: 234
  • Like
Reactions: xsf27 and alexi395

scum_bag

New Member
Nov 10, 2009
5
0
I'll not post more subs from https://sehuatang.org/ here, since people is spamming.

But you guys can create a thread and tag me and I'll try to help you

Edit: Looks like not everybody can create threads. So PM me and I will post the subs here
Hi man, sorry for the inconvenience, but i've been trying hard to lookup that sub, and signup the site, but it need invitation. that's why I try to post in the forum, maybe somebody can help to get that subs.

Thanks anyway :cheers:
 

Chuckie100

Well-Known Member
Sep 13, 2019
710
2,768
I found a machine (I think) translation for SPRD-1476. I cleaned it up a bit and attempted to correct several time slips on parts of the movie. Anyway enjoy and let me know what you think.
 

Attachments

  • SPRD-1476.rar
    8.8 KB · Views: 276

mei2

Well-Known Member
Dec 6, 2018
246
405
Here you go.

Good job picking up the dialogue over music -- I think it has done a much better job in that condition than Autosub.

As a comprison I ran autosub on the same movie. Here is the ja-JP srt result.

I also ran the NAVER translator on both versions: your aws version and the autosub -just to compare. It looks to me that autosub does a slightly better job in quiet or hiss noise. But AWS does much better job in noisy situatuion.

As a curiosity I also ran pyTranscriber on a portion of the video. I kind of thought pyTranscriber did a better job than autosub though it is supposed to be using autosub as its core engine :)
 

Attachments

  • SSIS-434-ja-JP autosub.zip
    9.3 KB · Views: 208
  • SSIS-434-en aws ja-Jp NAVER.zip
    11.2 KB · Views: 264

Electromog

Akiba Citizen
Dec 7, 2009
4,634
2,849
There were a few instances where it picked up words that weren't there, usually just single words like yes. Looks like it thinks *some* music is somebody speaking, but not most of it.
 

alexi395

Member
Jul 31, 2020
38
63
Good job picking up the dialogue over music -- I think it has done a much better job in that condition than Autosub.

As a comprison I ran autosub on the same movie. Here is the ja-JP srt result.

I also ran the NAVER translator on both versions: your aws version and the autosub -just to compare. It looks to me that autosub does a slightly better job in quiet or hiss noise. But AWS does much better job in noisy situatuion.

As a curiosity I also ran pyTranscriber on a portion of the video. I kind of thought pyTranscriber did a better job than autosub though it is supposed to be using autosub as its core engine :)
Thanks for the info, I observed AWS is much better for these kind of scenes (as in SSIS-434). Next I will run the audio through some filters and try to enhance the dialogues. Sometimes I think it will be much easier to actually learn Japanese. :rolleyes:


BTW @mei2 your filters for VSF are certainly better and do capture a lot more subs but it captures a lot of it partially . I mean one dialogue would span over 4 images due to the transition they have put . But thanks anyway I will try to tweak the other parameters when I get some time this weekend.
 

Chuckie100

Well-Known Member
Sep 13, 2019
710
2,768
I found a machine translation for HUNTA-632. I tried to clean it up and have posted it for your enjoyment and let me know what you think. OMG it was a 4 hr movie and I must have spent at least that much time just trying to clean it up. I can't even imagine doing one from scratch!
 

Attachments

  • HUNTA-632.rar
    22.9 KB · Views: 301

Pendekar2020

Well-Known Member
Sep 2, 2013
215
791
Thanks for the info, I observed AWS is much better for these kind of scenes (as in SSIS-434). Next I will run the audio through some filters and try to enhance the dialogues. Sometimes I think it will be much easier to actually learn Japanese. :rolleyes:

BTW @mei2 your filters for VSF are certainly better and do capture a lot more subs but it captures a lot of it partially . I mean one dialogue would span over 4 images due to the transition they have put . But thanks anyway I will try to tweak the other parameters when I get some time this weekend.
:D
Hang in there, I think speech recognition tech right now is already up to the task of creating a good subtitle. I think it's just the big techs (google, MS, Apple, Amazon) that intentionally put some limitations to it. The translator apps in smart phones have no problem translating (even in busy street) but suddenly got dumb down when it detect it's a movie.
 
  • Like
Reactions: mei2