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.
Sorry for asking but I DL and got this file JAV SRT Subtitle Cleaner & Paraphraser V1.0.ipynb. Could you attach or link the notes here?
[Reducing Mosaic]JUR-359 I Can’t Tell My Wife That I Got My Mother-in-law Pregnant… – On A Two-day, One-night Trip To A Hot Spring, I Lost Control And Kept Cumming Inside Her. – Iroha Morinaga
I couldn't resist subbing this recent release of a mother-son JAV. I used WhisperJAV0.7 to create this Sub and I also attempted to clean it up a bit and re-interpreted some of the meaningless/ "lewd-less" dialog. Again, I don't understand Japanese so my re-interpretations might not be totally accurate but I try to match what is happening in the scene. Anyway, enjoy and let me know what you think.
I wish "Mom" would have had a bush and a bit of a baby bump but I had to work with what I got!
Last edited:
No, you'd have to do that manually and import it. To do it automatically I'd have to figure out how to scrape that site, and given that it's a pay site, there might be built in mechanisms to prevent that. Maybe worth investigating for a future feature.
AVOP-381 - Summer Secret. Live Action Edition. (2017) [Muku]
This is my, almost perfect, .srt of AVOP-181 Summer Secret. Love Action Edition. It was done in an interesting way, because AVOP-181 is a word for word live adaption of the game by Otemotoya. So what I did was screen capture every screen in the game for the English translation text, and then used Photos app in Win 11 to "scan text" and copy and use that in Subtitle Edit.
A couple of lines I had to split across dialogues to match the flow, not sure if I missed a couple of screenshots, or extra dialogues lines were added to AVOP-181, but on the whole this is a 99.8% perfect subtitle. Enjoy!
This is my, almost perfect, .srt of AVOP-181 Summer Secret. Love Action Edition. It was done in an interesting way, because AVOP-181 is a word for word live adaption of the game by Otemotoya. So what I did was screen capture every screen in the game for the English translation text, and then used Photos app in Win 11 to "scan text" and copy and use that in Subtitle Edit.
A couple of lines I had to split across dialogues to match the flow, not sure if I missed a couple of screenshots, or extra dialogues lines were added to AVOP-181, but on the whole this is a 99.8% perfect subtitle. Enjoy!
Hello folks. I recently saw a post that alluded to my "clean_subs.py" file that is a post-processing script for my Whisper process. I assume it was mine since i arbitrarily chose that name. About Month or more ago I posted my Whisper configuration files but I have tinkered with them quite a bit to try to squeeze out the most accurate translation, including whispers and also advising Whisper that it is translating JAV adult content. As a result I have revised and revised and revised (You guys that are working with Whisper know what I mean) In any case I thought that anyone that is doing translations like these may benefit from some of my args (arguments-instructions) Please bear in mind that my stats may preclude some of the settings so my stats are: I9-14900K/Nvidia 4080 Super/96G RAM 6000mts/2TB NVMe for processing, and Windows 11 Pro. It particular the 4080 Super has 16G V-RAM so someone with 8Gig may need to tweak my settings a bit. One other little detail to pay attention to is that I am using Large-V3 and Faster-Whisper. That said, the following configs may contain some useful settings so here is what I use.
Last edited:
I upzipped it, how do I used your clean_subs.py and transcribe.py. Currently I'm using Faster-Whisper-XXL to subtitle English to get the English srt file. I would like to know how to use your python to clean up the srt. Thanks.
[Reducing Mosaic]RCT-402 An Incest Incident That Happened With A Miraculous Chance. A Mom Who Accidentally Felt Her Son’s Dick Inserted 2
I couldn't resist subbing this recent "reduced mosaic" release of an oldie Rocket mother-son themed JAV. I used WhisperJAV0.7 and my handy-dandy photo translator to create this Sub. I also attempted to clean it up a bit and re-interpreted some of the meaningless/ "lewd-less" dialog. Again, I don't understand Japanese so my re-interpretations might not be totally accurate but I try to match what is happening in the scene. Anyway, enjoy and let me know what you think.
I know there have been earlier subs for this video but had to try my hand at it since a reduced mosaic version came out recently.
[Reducing Mosaic]ALDN-464 Brainwashed Mother Lets Me Cum Inside Her Anytime Mikoshiba Mika
I couldn't resist subbing this recent "reduced mosaic" release of a mother-son themed JAV. I used WhisperJAV0.7 and I also attempted to clean it up a bit and re-interpreted some of the meaningless/ "lewd-less" dialog. Again, I don't understand Japanese so my re-interpretations might not be totally accurate but I try to match what is happening in the scene. Anyway, enjoy and let me know what you think.
I don't think I've seen a hypnotism themed JAV since the oldie Rocket RCT-322 father-daughter video! But after doing a search, I guess there were a bunch more!
Last edited:
So I ran the clean_subs.py with this cmd C:\Faster-Whisper-XXL\DSCOTTWHISPERCONFIGS>clean_subs.py hhd800.com@IQQQ-39.srt hhd800.com@IQQQ-39_clean.srt; the clean up is very subtle.
I couldn't get the transcribe.py to work. I got this error:
Traceback (most recent call last):
File "C:\Faster-Whisper-XXL\DSCOTTWHISPERCONFIGS\transcribe.py", line 4, in <module>
from faster_whisper import WhisperModel
ModuleNotFoundError: No module named 'faster_whisper'
Well I tried to formulate a cohesive response for you but my tangential babbling was characteristically all over the place so I employed Deepseek to help me to respond in a clear way for you, Here is the reply from Deepseek:
Response to Faster-Whisper-XXL Inquiry:
1. Setup Instructions
- Ensure you have Python installed (tested with Python 3.8+).
- Install the required dependencies:
bash
- pip install faster-whisper regex
- Place all three files (clean_subs.py, transcribe.py, and whisper.bat) in the same directory as your media files.
2. Usage Options
Option A: Automated Processing (Recommended)
- Simply double-click whisper.bat. This will:
- Transcribe all supported media files (MP4, AVI, MKV, etc.) in the directory using Faster-Whisper (large-v3 model by default).
- Automatically clean the generated .srt files using clean_subs.py.
Option B: Manual Cleaning (For Existing SRT Files)
If you already have an .srt file (e.g., from Faster-Whisper-XXL), run:bash
python clean_subs.py input.srt output.srt
This will apply the adult-content-optimized cleaning rules to your subtitles.
3. Notes on Faster-Whisper-XXL
- "XXL" is not an official Faster-Whisper model size. The standard options are:
- tiny, base, small, medium, large-v1, large-v2, large-v3.
- If your tool uses a custom model, ensure it outputs standard .srt files. The cleaning script will still work as long as the format is valid.
4. Customization
- To change the model (e.g., to large-v2), edit whisper.bat and update:
bat
- set "MODEL=large-v3"
- For language support, modify DEFAULT_LANG in the batch file or specify it manually:
bash
- python transcribe.py input.mp4 ja
5. Troubleshooting
- If errors occur, check whisper_results.log (for transcription) or the console output (for cleaning).
- Ensure your GPU drivers/CUDA are properly configured if using device="cuda".
One last little detail. If you do not want to use my Whisper.bat you can open each of these files in a wordprocessing program like word and compare each setting, such as the VAD perameters, and manually edit them. This is a good way to learn what each argument does, For example I have my settings at Float 16 but Float 32 will yield a more accurate translation at the cost of a much slower processing time. I'm using 8 CPU threads but my system is 24Core/32Thread. You may want to tinker with these settings depending upon your system. All of that said, My personal suggestion would be to make sure that Whisper is correctly pointing at your Faster-Whisper install directory but aside from that it should be a simple double-click process. Good luck.
Last edited:
Thanks for the clarification.
I guess my process can be like this
1. Run Fast-Whisper-XXL to get the English srt
2. And call your scrtipt clean_subs.py with this cmd C:\Faster-Whisper-XXL\DSCOTTWHISPERCONFIGS>clean_subs.py hhd800.com@IQQQ-39.srt hhd800.com@IQQQ-39_clean.srt;
I noticed the clean up script maily takes out the - , once in awhile uses lewd words and once in awhile take out repetative words. You find this true, like the diff above?
Hi Mycl500. I just edited that response while you were looking at it so you may gain some insight if you look at it again. The lewd words that you allude to are easily edited if you use word to open the file, remove the entire line if you do not like a particular translation or if you want to add your own you can use the same structure as what I have there. Another item that many people don't know is that you can edit these finished .srt files with a word program so if you use, for example OpenOffice you can open the file in openoffice and then use the find and replace option to remove certain words or change them. YOu could, for example, change all instances of "cum" and do a Find-Cum Replace with BubbleGum, replace all, and when you play your video, ta-ta, now the girls are anxious to swallow a load of bubblegum. IN any case, take a look at my Deepseek created directions, I hope that helps.
No i want more lewd words....I was wondering why japanese porn doesn't used lewd words, now I realized it Whisper translation that's replacing it with bubblegum :-(
Bubblegum?
I've wondered the same thing you have, about why JAV doesn't use more lewd words.
My best guess is that they are following the whole "less is more" philosophy and trying to avoid the words losing their weight.
It's just a guess though.
As I was contriving various word substitudions my 'a'int I a stinker' mindset was always coming up with weird and wonderful sustitutions like replacing ramen with' old men in funny hats' or a wide variety of stupidity but since I was doing these subs mainly for the akiba cats I decided to go with the obvious. I think that once people realize that they can easily edit these .srt files that eventually you'll start to find all kinds of creative insanity. I hope so at least.
Ok thank you for the instructions above:
To get the process to work for me:
1. I had to install: pip install faster-whisper regex
2. I had to install cuDNN 9.11.0: cudnn_9.11.0_windows from https://developer.nvidia.com/cudnn-downloads
a. I had to put this in my window path: C:\Program Files\NVIDIA\CUDNN\v9.11\bin\12.9
3. I had to install CUDA Toolkit 12.9: cuda_12.9.1_576.57_windows from https://developer.nvidia.com/cuda-d...86_64&target_version=11&target_type=exe_local
Attached is
1. Your process srt: DScott_hhd800.com@WAAA-194.srt = 44 kb
2. My process srt using fast-whisper-xxl and then using just your clean_subs.py to clean up my fast-whisper-xxl: mycl500_hhd800.com@WAAA-194.srt = 72 kb using --model large-v2
3. Also you can see the diff of the 2 files above.....crazy how everything is so different but context is similar. I guess I will have to play around and see which process I like bettter. Both process of creating the srt are fast like 5-10 mins depends on the movie length.
Thanks DScott....now I have another weapon to use.
That is excellent news. Please keep us posted if you find a superior configuration. My results are mostly a battle with accuracy vs. hallucinations. Ultimately I want the best accuracy but I have made some compromises to try to avoid those damned hallucinations.
IF you discover the ultimate configuration, please let me know and if it yields a superior result to my recent mass-srt adventure, I will do it again with the better settings.
It was very interesting the discrepancies between the Two files. One thing that I try to minimize with my configurations is to prevent Whisper from 'assuming' text'. I don't want it to inject what it 'guesses' that is there or fills in phrases, like thank you for subscribing, because there is a gap in the dialog. I don't want AI to guess I want it to translate exactly what it 'hears' and not another word extra.
I'm glad that you were able to get it working. Cheerss.
Last edited by a moderator: