Whisper (OpenAI) - Automatic English Subtitles for Any Film in Any Language - An Intro & Guide to Subtitling JAV

[montblanc5]

I've been getting a crash error my most recent time running. My files are in the drive folder as normal, no other changes have been made.

Edit: Nevermind, my bookmark was for an older version. Everything works fine with the most up to date WhisperJAV.

 

[ArtemisINFJ]

Credit goes to @anon_entity, the original developer of WhisperVAD. He had put together a robust algorithm.
I was not planning any update to that colab. But if there is any ideas for improvments or new features, I'm happy to look into it.

I think you can add some feature to clean up the transcriptions afterwards since it takes some time to do post editing the subs with the amount of hallucinations it produces specifically with moaning sounds and etc.. looking forward to see if it possible for you to add that feature. Thanks.

 

[JavNoob]

Does anybody know how to fix this ?
I am using WhisperJAV_v0_7a.ipynb in browser.

 

[mei2]

Does anybody know how to fix this ?
I am using WhisperJAV_v0_7a.ipynb in browser.

It might have been caused by version incompatibility. Try WhisperJAV_v0_7b version.

 

[JavNoob]

It might have been caused by version incompatibility. Try WhisperJAV_v0_7b version.

You are correct. Thank you very much.

 

[Elmerbs]

I've been using openAI small model because my crappy 6GB GPU. It would take around 40mn to 1h30 for any single movie. It was very discouraging.
Update: I've finally found time to make a build of whisper.cpp with cuda support.
After a little fine tuning of the params, a 2h movie takes 5mn with the medium model!
Direct translation looks even better and the only post processing are just dup lines and none of the gibrish I used to get before like "ahhhhhhhhh" "I like it that way.I like it that way.I like it that way.I like it that way.I like it that way.", etc...
Just clean text that makes sense.
Total game changer for me.

 

[Imscully]

I've been using openAI small model because my crappy 6GB GPU. It would take around 40mn to 1h30 for any single movie. It was very discouraging.
Update: I've finally found time to make a build of whisper.cpp with cuda support.
After a little fine tuning of the params, a 2h movie takes 5mn with the medium model!
Direct translation looks even better and the only post processing are just dup lines and none of the gibrish I used to get before like "ahhhhhhhhh" "I like it that way.I like it that way.I like it that way.I like it that way.I like it that way.", etc...
Just clean text that makes sense.
Total game changer for me.

This line of dialog almost made me wonder if it was a take on the old song lyrics "Oooo, Baby....I like it like that!"
Was it making subs form a JAV or old MTV video?

 

[DocNic]

Hello all, I'm getting the following error when using Whisper.



I've tried all the suggestions Gemini has offered but none seem to work, it seems the Pytorch and Torchdec are not compatible. Any suggestions or can you point me to a copy of Whisper that currently works? Thank you!

 

[mei2]

@DocNic
This one:

WhisperJAV/notebook/WhisperJAV_colab_edition_expert.ipynb at main · meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV - meizhong986/WhisperJAV

github.com

You can use the default settings --they are already tuned for most jav genres. But if you want to get more out of it, please look into the readme.md file in github. For example change the pipeline to fidelity to get the best quality with large-v2 model.

I'm planning to look into whisperwithvad_pro fo fix it during the early part of the week when I get some bandwitdh.

 

[DocNic]

@DocNic
This one:

WhisperJAV/notebook/WhisperJAV_colab_edition_expert.ipynb at main · meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV - meizhong986/WhisperJAV

github.com

You can use the default settings --they are already tuned for most jav genres. But if you want to get more out of it, please look into the readme.md file in github. For example change the pipeline to fidelity to get the best quality with large-v2 model.

I'm planning to look into whisperwithvad_pro fo fix it during the early part of the week when I get some bandwitdh.

Thanks mei2! Really appreciate it!

 

[mei2]

Hello all, I'm getting the following error when using Whisper.

I've tried all the suggestions Gemini has offered but none seem to work, it seems the Pytorch and Torchdec are not compatible. Any suggestions or can you point me to a copy of Whisper that currently works? Thank you!

I just updated WhisperWithVAD_pro to fix the issue. I don't have the bandwidth to test it right now. Please let me know if the issue persists.
My take was a bit different than gemini. The real issue is not Torchdec. The root cause seems to be that silero has updated its version and uses pydub which in return needs a backend for audio. By default pydub wants to use Torchdec, but that is not compatible with colab stack. Instead I have added soundfile as the backend. Soundfile is less problematic and more robust. So it should work: silero -> pydub -> soundfile .

If any issues let me know. BTW the link to the updated version is the same place:

WhisperJAV/notebook/WhisperWithVAD_pro.ipynb at main · meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV - meizhong986/WhisperJAV

github.com

 

[DocNic]

I just updated WhisperWithVAD_pro to fix the issue. I don't have the bandwidth to test it right now. Please let me know if the issue persists.
My take was a bit different than gemini. The real issue is not Torchdec. The root cause seems to be that silero has updated its version and uses pydub which in return needs a backend for audio. By default pydub wants to use Torchdec, but that is not compatible with colab stack. Instead I have added soundfile as the backend. Soundfile is less problematic and more robust. So it should work: silero -> pydub -> soundfile .

If any issues let me know. BTW the link to the updated version is the same place:

WhisperJAV/notebook/WhisperWithVAD_pro.ipynb at main · meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV - meizhong986/WhisperJAV

github.com

Awesome, I'll give it a try!

 

[DocNic]

Awesome, I'll give it a try!

Hello mei2, it's still giving the Torchcodec error: I went with the Gemini suggestions and but it could never resolve the Torchcodec and Pytorch compatability issue.

 

[mei2]

Hello mei2, it's still giving the Torchcodec error: I went with the Gemini suggestions and but it could never resolve the Torchcodec and Pytorch compatability issue.

Damn, just when I thought I found the right workaround
I'll look into it during the weekend.

 

[mei2]

Hello mei2, it's still giving the Torchcodec error: I went with the Gemini suggestions and but it could never resolve the Torchcodec and Pytorch compatability issue.

I have updated the notebook: WhisperWithVAD-PRO. I tested it with a single audio and it worked fine. Let me know if the error resurfaces.

WhisperJAV/notebook/WhisperWithVAD_pro.ipynb at main · meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV - meizhong986/WhisperJAV

github.com

 

[ToastFrench]

I have updated the notebook: WhisperWithVAD-PRO. I tested it with a single audio and it worked fine. Let me know if the error resurfaces.

WhisperJAV/notebook/WhisperWithVAD_pro.ipynb at main · meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV - meizhong986/WhisperJAV

github.com

Hi Mei,
sorry for the completely clueless question (I don't really understand GitHub or compiling, lol) but for an Apple Silicon Mac, I just copy the lines in your Release WhisperJAV v1.8.1 installer instructions into Terminal, Enable Local LLM Translation, and then "whisperjav-gui" and then go from there?

(I want to try your main WhisperJAV, not the with VAD-Pro etc..dependencies, thanks!)

 

[mei2]

Hi Mei,
sorry for the completely clueless question (I don't really understand GitHub or compiling, lol) but for an Apple Silicon Mac, I just copy the lines in your Release WhisperJAV v1.8.1 installer instructions into Terminal, Enable Local LLM Translation, and then "whisperjav-gui" and then go from there?

(I want to try your main WhisperJAV, not the with VAD-Pro etc..dependencies, thanks!)

For Apple Silicon I suggest you use the full version. The full version releases are listed here:

Releases · meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV - meizhong986/WhisperJAV

github.com

The latest release is v1.8.2-rc02. See the section for Apple Silicon installation:

git clone https://github.com/meizhong986/whisperjav.git
cd whisperjav
git checkout v1.8.2-rc2
macOS: brew deps + ./installer/install_linux.sh


I don't have a Mac to test all the options. The community is probably more helpful:

meizhong986 WhisperJAV · Discussions

Explore the GitHub Discussions forum for meizhong986 WhisperJAV. Discuss code, ask questions & collaborate with the developer community.

github.com

Also, one of the users recently posted that they have gotten the GUI to work on MacOS. The GUI makes your life easy and it has more features than CLI.

 

[DocNic]

I have updated the notebook: WhisperWithVAD-PRO. I tested it with a single audio and it worked fine. Let me know if the error resurfaces.

WhisperJAV/notebook/WhisperWithVAD_pro.ipynb at main · meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV - meizhong986/WhisperJAV

github.com

Just tested, works fine! Thank you for all your hard work, really appreciate it!!

 

[ToastFrench]

Hi mei2,
after seeing another user running this successfully with an M1 Mac, I gave it another try, and keep getting stuck here:

[13:50:53] [ERROR] Install core dependencies - failed after 3 attempts
Traceback (most recent call last):
File "/Users/toasted/whisperjav/install.py", line 1689, in <module>
main()
File "/Users/toasted/whisperjav/install.py", line 1252, in main
run_pip(executor, ["install"] + core_deps, "Install core dependencies")
File "/Users/toasted/whisperjav/install.py", line 1020, in run_pip
result = subprocess.run(
^^^^^^^^^^^^^^^
File "/Users/toasted/.pyenv/versions/3.12.12/lib/python3.12/subprocess.py", line 571, in run
raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['/Users/toasted/venvs/whisperjav/bin/python3', '-m', 'pip', 'install', 'faster-whisper>=1.1.0', 'tqdm', 'colorama', 'requests', 'aiofiles', 'regex', 'pysrt', 'srt', 'pydantic>=2.0,<3.0', 'PyYAML>=6.0', 'jsonschema', 'numpy>=1.26.0,<2.0', 'scipy>=1.10.1', 'numba>=0.58.0', 'soundfile', 'pydub', 'librosa>=0.10.0', 'pyloudnorm', 'auditok', 'silero-vad>=6.2', 'ten-vad', 'psutil>=5.9.0', 'scikit-learn>=1.3.0', 'av>=13.0.0', 'imageio>=2.31.0', 'imageio-ffmpeg>=0.4.9', 'httpx>=0.27.0', 'websockets>=13.0', 'soxr>=0.3.0']' returned non-zero exit status 1.

EDIT: I fixed it: For Mac users using Brew, you might need to run:

brew install portaudio
pip install pyaudio

I see the GUI!

 

[SamKook]

Seems odd to me that you have 2 different python folder involved(if you use a virtual environment, the system python won't see installed modules on it for example and vice-versa), but I don't know how his installation works at all so that might be normal.