I used .5, .6, and .8. The .8 had the most lines, but was off by up to 30-seconds timewise. It wasn't a good test movie as there was a lot of mumbling. I'm going to try something else a little clearer.
Post your JAV subtitle files here - JAV Subtitle Repository (JSP)★NOT A SUB REQUEST THREAD★
- Thread starter Eastboyza
- Start date
-
Akiba-Online is sponsored by FileJoker.
FileJoker is a required filehost for all new posts and content replies in the Direct Downloads subforums.
Failure to include FileJoker links for Direct Download posts will result in deletion of your posts or worse.
For more information see this thread.
That's odd. I just tried processing a file at .6 then .2, and the .2 gave more lines. One would think if it is a detection threshold for audio, the lower number would catch more.
The problem with comparing setting is that whisper isn't reliable. Even using the exact same settings, you'll get a different number of lines every time(never tested the same audio more than 4 times though but that's a lot already).
No idea how it can sometimes fail and sometimes not with the same everything but it does.
Wow, that is extremely strange. If the code remains the same, it should return the same result with the same input every time.
Beginning to wonder if I should just try re-running those that didn't turn out well and see if they randomly get better.
May i know the average size of an audio that is extracted from 1-hour video? Because when i am converting a 1-hour video to mp3 it has an average size of 120mb. I wonder what is the size of an extracted audio compared to mp3. Thank you for your explanation btw.
May i know the average size of an audio that is extracted from 1-hour video? Because when i am converting a 1-hour video to mp3 it has an average size of 120mb. I wonder what is the size of an extracted audio compared to mp3. Thank you for your explanation btw.
Depends entirely on the bitrate used to encode the original audio. What matters most is the quality of the audio. With equal size, between mp3, m4a(aac) and opus, opus will sound closer to the original, followed by aac, followed by mp3.
And if you encode the original audio to either of those 3 format, you're going to lose some quality which could affect the ability of whisper to recognize voice(but in practice it doesn't seem to have much effect, although it's hard to tell with the randomness of it).
That's why I recommend keeping the audio as-is and just separating it from the video, you get the best possible quality this way and as a bonus, the size is usually smaller than if you use quality settings for mp3s.
To answer your question more directly, these days, I see either 192Kbps aac audio in most HD releases which would be about 87MB per hour if I did my math right or 128Kbps which would be 54MB and also the most common bitrates you'll encounter for any audio from ripped movies.
Sounds like you're using 256Kbps for your mp3s which would be around 115MB with the same math which is bigger than most audio I've seen, it rarely goes above 192Kbps(for ripped content, DVD/BluRay audio is much bigger).
I use 128Kbps opus for my own encodes for the quality version(which is of similar quality than your mp3 but only half the size) and 64Kbps for the small version.
I used mp3 in the beginning, changed to mp4 and now wav. Whatever input you give, it converts it to wav mono.
It also uses ffmpeg to load the audio so it'll support just about anything, what's in the audio path is just an example.
It gets converted internally in whisper to that either way so all whisper processes is mono audio. For those other software, maybe they try to do extra stuff before whisper and they don't like mono or you're converting to mono in a bad way and it messes up the quality real bad which is why it has issues, although I couldn't say how you'd achieve that other than converting to 16k on stereo first and then 16k mono maybe.
It seems pretty random. I would miss a completely clear line on a .4 than a .8. But overall it is a pretty sweet program. Sometimes only a minor language cleaning is needed. I will never complain about the quality for the cost.
First, I really want to thank SamKook, soloporhoy666 and others for posting info about how to use Whisper collab. I finally got around to trying it, got a few videos done and hit the limit. I logged into a new gmail account but still get an error, though it is different from the limit-reached error message. Two questions, one, do you know how long the limit-reached error is in effect, like does it reset after 24 hrs or so? And two, the 2nd error was a time out error, but it came up almost right away. When switching to a new gmail account, perhaps I should have restarted Chrome? Any other suggestions for managing this limit for free users? Thank you again to all posters for helping with Whisper. I will add the srt files I generated here in case anyone else wants them. I have not cleaned them up though. If I do, I will repost if anyone wants them, just let me know. The titles in this archive are
DMOW-163
MIAA-633
QRDA-151
XRLE-002
DMOW-163
MIAA-633
QRDA-151
XRLE-002
So I have been messing around with Whisper for a while now and as a result have produced a decent amount of un-edited sub files that I probably will never clean up alongside some I do plan to clean and post here. I was initially not planning on posting those un-edited ones here but if there was any interest in those I would be down to post them. As per the ones I plan to clean up I still have my original project of cleaning up Real-674 that I found on SubtitleCat that I have been using Whisper to help me out as another reference and have also decided to work on cleaning the Whisper file for MISM-165 since its probably my favorite deepthroat JAV and there are barely any subs for deepthroat videos out there. It hopefully should not be that bad since despite its length its pretty light on unique dialogue.
There's very little skill involved into doing it yourself so the least you could do is try. All the common problems you could run into have a solution on this thread as well as multiple tutorials on how to do it.
If you run into an issue, just post your detailed problem here and people will help you out.
More Whisper subs:
ASW-098
DJE-051
HDD-001
HDD-008
HDD-010
KV-075
KV-201
KV-226
KV-227
KV-228
KV-243
KV-247
MIGD-680
ASW-098
DJE-051
HDD-001
HDD-008
HDD-010
KV-075
KV-201
KV-226
KV-227
KV-228
KV-243
KV-247
MIGD-680
I don't want to sound harsh. But I have a question.
Why keep creating "un-edited sub files" and not edit the ones you already have?
Why don't you slow down and clean up the ones you already have?
Wouldn't it be easier to ask others what they would like subbed?
Don't get overwhelm. As I sub my own, I understand. There's lots of movies out there.
Just slow down and take your time. The amount of time you spend with your subs is what gonna make it the best for you
Last edited:
Haha no worries about sounding harsh. The reason I have all these un-edited files is a combination of me messing around and getting used to the program and making files just for my personal use. The way I approach subbing is just doing the bare minimum for videos I only somewhat like or haven't watched yet while videos I have already seen and really like I go back and clean up those sub files. I assume people would only be interested in cleaned up files but wasn't sure which is why I asked.I don't want to sound harsh. But I have a question.
Why keep creating "un-edited sub files" and not edit the ones you already have?
Why don't you slow down and clean up the ones you already have?
Wouldn't it be easier to ask others what they would like subbed?
Don't get overwhelm. As I sub my own, I understand. There's lots of movies out there.
Just slow down and take your time. The amount of time you spend with your subs is what gonna make it the best for you
I use 2 accounts, the limit I have is 6 files before the program gives me an error, I upload audio files of 200 mb 1x1 first I upload one, then during the transcription process I upload the other, recently I have detected that whisper now play with the use of ram memory, which appears in the upper right part, if the box exceeds the use limit, the whisper program will also close, so now you have to pay attention not only with the use of the GPUs as well In the use of ram memory, I also recommend refreshing the file folder, so that the page does not fail, if you are using more pages, it constantly returns to whisper and refreshes, so that it seems that you are active on the page.
