JAV WHISPER+VAD Request Thread (Taking Requests HERE)

summerss

Member
Jun 8, 2009
43
11
I see. Seems the biggest issue is that people get confused between the different versions available, even though it's all basically the same thing but with some tweaks.

If you say whisper + VAD, people will only think you're talking about the google colab thing(or a local clone of that) since that's 2 separate software working together, not just whisper and the easy UI for that is the colab webpage. You also have a pretty much exact tutorial on how to install it on linux if you just press show code on the relevant steps.
I think i see now. So this webpage google colab

already has the whisper+Vad, and its not something i install on PC unless i have linux (i'm on windows)? and its not something i add on to subtitleedit. Subtitle edit is for fixing stuff after?
 

SamKook

Grand Wizard
Staff member
Super Moderator
Uploader
May 10, 2009
3,755
5,154
Yeah, the webpage is all you need. It installs everything on a remote computer and allows you to use it(with some limitation as a free user in the form of a certain amount/time of gpu use per a period of time).

You can also replicate that installation on your own pc to use it without limitation but it'll be slower unless you have a good AI GPU. It'll work on windows but the colab is on linux so the commands to install everything might vary slightly for windows.

SubtitleEdit has its own audio to text options, one of which uses whisper(it'll download and install what you need), but people mostly use it to edit subtitles created from the colab since whisper on its own has issues with japanese, which the addition of a VAD helps alleviate and it's far from perfect so manual editing of some kind is required to achieve a better result.
 

summerss

Member
Jun 8, 2009
43
11
Yeah, the webpage is all you need. It installs everything on a remote computer and allows you to use it(with some limitation as a free user in the form of a certain amount/time of gpu use per a period of time).

You can also replicate that installation on your own pc to use it without limitation but it'll be slower unless you have a good AI GPU. It'll work on windows but the colab is on linux so the commands to install everything might vary slightly for windows.

SubtitleEdit has its own audio to text options, one of which uses whisper(it'll download and install what you need), but people mostly use it to edit subtitles created from the colab since whisper on its own has issues with japanese, which the addition of a VAD helps alleviate and it's far from perfect so manual editing of some kind is required to achieve a better result.
Cool, thx. I tried just pressing play on everything and uploading an english file first.
Do you have concerns about privacy? I see the words Colab and i'm like, are others seeing this? is it deleting everything on close/refresh?

OK i think i found the instructions to run locally are these correct
Still looks intimidating as hell.

Ill stop asking questions if you are already planning to make your own tutorial for all this later on. I can wait for that.
 

SamKook

Grand Wizard
Staff member
Super Moderator
Uploader
May 10, 2009
3,755
5,154
There is a privacy concern if you link your google drive to it without making sure the code doesn't do anything funny with it since you give it full access. You have the option to upload directly to the session on the left which is what I do, but that might mean having to upload the file more than once if there's an issue.
Others don't see what you do(except google obviously) and it all get deleted(if you believe google) when you disconnect the session so not much to worry about(from anyone but google itself).

Those instruction are probably fine for one way to do it, but there are easier ways. I personally find using the command line version easier than installing a web ui for it but I'm very familiar with that kind of stuff.

I am writing something in the tutorial section atm Done writing something in the tutorial section. I'll probably post it in an incomplete form today(since I want people to be able to ask questions there) and add more detailed instructions on how to use and install different versions for it later. I'll try to at least add how to use the web version today.
 
Last edited:

asledan

Member
Feb 20, 2016
59
97
I have a request, but it's really to help me with a problem I've been having with whisper. I use Whisper with VAD on google colab and at first it was working fine, but now more and more often my subtitles tat to drift ou of sync as the film progresses until by th end of he film the subtiles are completely out of sync. I am extracting the audi fom teh video fles using avidmux and uploading the aac file to google drive, thne run it through the colab. inally I use Deeplv4 to translate the srt to Englsih.

My latest attempt is Ebod-993 with a blond Arai Rima!



Can you use the same source file and whisper it to se if you get better results to help me figure out whats going wrong on mine end?
The film is here:

and here are my drifty subtitles and he aac file...


View attachment 3323484

I had this one done but then I realized I did it with default settings instead of my usual setting recommendations, so Im redoing and posting tmmr.

Everyone else! Here's a partial dump. I'm still working on a 2-3 titles but the majority are here. I went through all the requests but I could only do so many based on what I could find source videos for and what I might've wanted for my own collection. Some requests were part of the 28000+ chinese subtitle dump, so those were easy to do. As this is a free service I cannot guarantee I get to all the requests so apologies if I didn't get to you.
 

Attachments

  • Sub dump 9.24.23.zip
    179.4 KB · Views: 245

asledan

Member
Feb 20, 2016
59
97
Thanks for the extensive evaluation, I looked at all of the different subtitle files you created and I see the same subtitle drift I was having, it's most notible near the end of tehfilmm as it is an additive effect.

Look at the following screen grab, subtitle number 977 "hello" does not belong at 1:55:51.750in this post coital scene (even though the audirogram shows it matching mushi muhi!")....


View attachment 3324109


it belong at 01:56:23, shown in the grab below as the guy answer the phone! (and the audiograph doesn't match "mushi mushi"!) The subs are now 32 seconds out of sync!



View attachment 3324110



I think the audiograph is showing that he audio and video are out of sync... any thoughts?
I finished my own run through of whisper+vad for EBOD-993 if you needed another comparison. As well I finished another requested translation so they are both in the folder.
 

Attachments

  • subs.zip
    39.1 KB · Views: 159

asledan

Member
Feb 20, 2016
59
97
Yeah, the webpage is all you need. It installs everything on a remote computer and allows you to use it(with some limitation as a free user in the form of a certain amount/time of gpu use per a period of time).

You can also replicate that installation on your own pc to use it without limitation but it'll be slower unless you have a good AI GPU. It'll work on windows but the colab is on linux so the commands to install everything might vary slightly for windows.

SubtitleEdit has its own audio to text options, one of which uses whisper(it'll download and install what you need), but people mostly use it to edit subtitles created from the colab since whisper on its own has issues with japanese, which the addition of a VAD helps alleviate and it's far from perfect so manual editing of some kind is required to achieve a better result.
Any advice on how to replicate whisper+vad on pc? I have very minimal coding/python experience so I do what I can with the google collab. I actually have a solid setup I think might work well.
 

SamKook

Grand Wizard
Staff member
Super Moderator
Uploader
May 10, 2009
3,755
5,154
Any advice on how to replicate whisper+vad on pc? I have very minimal coding/python experience so I do what I can with the google collab. I actually have a solid setup I think might work well.
If you look at the code on the colab(by pressing "show code") and adapt it for windows, you can replicate it on your pc. It's mostly commands to run to install/run stuff and a python script at the end(I think). Might have to install a few extra things for stuff that might be included on linux but not windows by default, but looking at the errors you get will tell you if that's required.

I do plan to add how to do that in the whisper tutorial I made eventually but I have other priorities at the moment.
 

bathsheba666

New Member
Sep 18, 2023
8
1
I had this one done but then I realized I did it with default settings instead of my usual setting recommendations, so Im redoing and posting tmmr.

Everyone else! Here's a partial dump. I'm still working on a 2-3 titles but the majority are here. I went through all the requests but I could only do so many based on what I could find source videos for and what I might've wanted for my own collection. Some requests were part of the 28000+ chinese subtitle dump, so those were easy to do. As this is a free service I cannot guarantee I get to all the requests so apologies if I didn't get to you.
Many many thanks for these subs. I spent quite a few years wondering what exactly was cajoling or coercion.

One nit - ipz-288 seems to be missing the first almost one hour of text ?

Really happy you got Emily and Rio.
 

avatarthe

Well-Known Member
Feb 1, 2008
184
282
I finished my own run through of whisper+vad for EBOD-993 if you needed another comparison. As well I finished another requested translation so they are both in the folder.
Thanks! Your subtitles confirm my hypothesis that the issue is subtitles created with whisper don’t work with movies encoded with a variable frame rate, even it for use the VFR film as the source the resulting subtitles will drift out of sync with the VFR film but will work with a copy of the movie encoded with CFR
 
Last edited:

pavane

Active Member
Feb 8, 2010
196
47
Hi
thanks for the subtitles to RBD 303,940 and 838.
I couldnt reply sooner as I was in a no net area for 7 days!
Thanks again.
 

pukimai

Active Member
Jan 11, 2013
244
54
hi, movies type that i really curious and wants to have the subtitle to know the context is actress debut, so it's possible to provide subtitle for this debuts movies? or in the meantime i also try to learn how to use whisper using my potato PC:
1. XV-646 - Itou Haruka
2. AVOP-201 - Aizawa Minami - available in sextb
3. SNIS-190 - Amatsuka Moe
4. IPZZ-166 - An Coco
5. MAS-086 - Harumiya Kokoro
6. SNIS-632 - Hashimoto Arina
7. STAR-888 - Honjou Suzu
8. MIDE-974 - Ishikawa Mio
9. SSNI-064 - Kano Yura
10.SSIS-115 - Kasui Jun
11.MIDV-075 - Miyashita Rena
12. IPZ-294 - Miyazaki Rinon
13.SSIS-736 - Mochizuki Tsubomi
14.STARS-97 - Nagano Ichika
15.MIDE-488 - Nanasawa Mia
16.EDD-218 - Suzumura Airi
17.TEK-096 - Sakura Moko
18.STAR-828 - Takeda Yume
19.STAR-840 - Takeda Yume
20.SDAB-014 - Toda Makoto
21.BGN-015 - Yatabe Kazuha
22.ONED-313 - Sakura Saki

Apologize that i have so many in pipeline, i think better for me to learn how to use whisper and do it by myself, again apologize
 
Last edited:

pavane

Active Member
Feb 8, 2010
196
47
hi
have just been viewing some Jav subs you did for me.
RBD-303 subs are incomplete they stop at about 1 hr 57 minutes.
I think this is because all the downloads I can find are in 2 parts and
the first half stops at that time.
So could you please do all of RBD-303?
 

Pein1911

Active Member
Sep 20, 2019
41
127
hi, movies type that i really curious and wants to have the subtitle to know the context is actress debut, so it's possible to provide subtitle for this debuts movies? or in the meantime i also try to learn how to use whisper using my potato PC:
1. XV-646 - Itou Haruka
2. AVOP-201 - Aizawa Minami - available in sextb
3. SNIS-190 - Amatsuka Moe
4. IPZZ-166 - An Coco
5. MAS-086 - Harumiya Kokoro
6. SNIS-632 - Hashimoto Arina
7. STAR-888 - Honjou Suzu
8. MIDE-974 - Ishikawa Mio
9. SSNI-064 - Kano Yura
10.SSIS-115 - Kasui Jun
11.MIDV-075 - Miyashita Rena
12. IPZ-294 - Miyazaki Rinon
13.SSIS-736 - Mochizuki Tsubomi
14.STARS-97 - Nagano Ichika
15.MIDE-488 - Nanasawa Mia
16.EDD-218 - Suzumura Airi
17.TEK-096 - Sakura Moko
18.STAR-828 - Takeda Yume
19.STAR-840 - Takeda Yume
20.SDAB-014 - Toda Makoto
21.BGN-015 - Yatabe Kazuha
22.ONED-313 - Sakura Saki

Apologize that i have so many in pipeline, i think better for me to learn how to use whisper and do it by myself, again apologize
You can run it with a potato pc :)), Collab will save you a alot, because its running on cloud base. If you choose to run it yourself on your computer, it will take a lot of time, as well as electricity costs. Good luck
 

pukimai

Active Member
Jan 11, 2013
244
54
tried using google collab, really take time for uploading the movie ya.... hahahah
 

Goodsky

New Member
Aug 31, 2022
8
8
hi, movies type that i really curious and wants to have the subtitle to know the context is actress debut, so it's possible to provide subtitle for this debuts movies? or in the meantime i also try to learn how to use whisper using my potato PC:
1. XV-646 - Itou Haruka
2. AVOP-201 - Aizawa Minami - available in sextb
3. SNIS-190 - Amatsuka Moe
4. IPZZ-166 - An Coco
5. MAS-086 - Harumiya Kokoro
6. SNIS-632 - Hashimoto Arina
7. STAR-888 - Honjou Suzu
8. MIDE-974 - Ishikawa Mio
9. SSNI-064 - Kano Yura
10.SSIS-115 - Kasui Jun
11.MIDV-075 - Miyashita Rena
12. IPZ-294 - Miyazaki Rinon
13.SSIS-736 - Mochizuki Tsubomi
14.STARS-97 - Nagano Ichika
15.MIDE-488 - Nanasawa Mia
16.EDD-218 - Suzumura Airi
17.TEK-096 - Sakura Moko
18.STAR-828 - Takeda Yume
19.STAR-840 - Takeda Yume
20.SDAB-014 - Toda Makoto
21.BGN-015 - Yatabe Kazuha
22.ONED-313 - Sakura Saki

Apologize that i have so many in pipeline, i think better for me to learn how to use whisper and do it by myself, again apologize
A lot of these are available at subtitlecat.com