Search results

Nice, I did the same. Sure, please feel free.

Need to try it but deepseek usually has an issue with sending the whole file at once. Token max is reached pretty soon.

They are really slow. One line of sub takes generelly 1.5s.

I played around with deepseek and VSM and have no idea how it works for you. For me, half the subtitle is not translated and the other half is empty. Therefore, I wrote my own little script which does the exact same thing. I also tried one which works with chunks instead of each line seperately...

Tried deepseek (free version) and the results are way better than deepl. I see the tokens are cheaper than ChatGPT but what is your usual cost per subitle file as a reference?

Really? Even though you loose the explicit language?

Have you tried using deepl? If yes, how does it compare to Deepseek?

Just drag it into your GDrive, double click and it opens as the normal whisper script would. Then just adjust the one line I mentioned in my last post

I downloaded my adapted script. Just upload it to your colab. Works slightly different. Make sure to change this line in the code to the path where you store the audio: audio_folder = "/content/drive/MyDrive/Colab Notebooks/" It will pick the audio from there and run it.

This one: https://colab.research.google.com/github/meizhong986/WhisperJAV/blob/main/notebook/WhisperWithVAD_pro.ipynb#scrollTo=sos9vsxPkIN7

These are two vastly different whisper implementations. Whisperjav 0.7 has a lot more clean-up built in, which makes the immediate results seem better but the overall precision is lower due to the use of faster-whisper. Yes, it is faster and more cleaned right out of the colab but quality wise...

Yes, large-v2. Was also using large-v1 and large-v3 but all of them have the same flaw. There seems no apparent reason. Sometimes it is perfectly fine but then completely out of sync. Even retrying does not solve the issue.

Does anybody else experiences timing issues fairly often when using WhisperWithVAD (up to ~30sec sometimes)?

Yeah, had the same observation. Seems Whisper is really strict with the package range they allow and manually forcing other versions results only in more/other issues.

I think this is code he's using: https://colab.research.google.com/github/ANonEntity/WhisperWithVAD/blob/main/WhisperWithVAD.ipynb

I'm using standard settings and the large model from the collab posted here very early on. (VAD threshold 0.4 and chunk_thershold 0.3)

I did one file today with only transcribe and it took nearly 1h 30min instad of approx. 30min, then ran it through Deepl and I wasn't impressed with the result. Did not seem better than just straight letting whisper do the whole job.

This will not work as Whisper generates different subtiltes every time you run it, so there is no one-to-one comparison unless you get the japanese version with no translation and the translated one from the same run.

I know how it workes, the question is if you get better results using whisper or deepl.

Has anybody tested how whisper translation compares to deepl (free version)? Up to know I found whisper doing a very good job but since I can't save a japanese version for deepl and let whisper translate it for he same file, it's hard to tell.

Sounds interesting. Do you know how the command should look like for command line users like me?

Thanks for looking into it. That was exactly the issue. It merges the file, but combines the last 2 rows of a chunk together, so the whole thing is not usable. Thanks anyways :) Edit: Was able to fix it after 3h of reverse engineering. Man was that a pain but was worth the effort. Your...

Thanks a lot! Everything seems to be working but deepl changed the input css selector, which I fixed but then for some reason it messed with the merge of the chunks into one file. All srt which only have one chunk work, but as soon as more than one is required, it fails.

If anyone is intersted, I have a large amount of Whisper generated subtitles. They are NOT cleaned, because I am only interested in understanding the story, not having perfect subs. Duplicates are removed as good as possible and timings slightly fixed.

Is anyone using a script do auto sub srt with deepl (free version)? Mine broke and I can't seem to fix it so I am looking for an alternative or someone more technical than me who knows how to fix it.

I used mp3 in the beginning, changed to mp4 and now wav. Whatever input you give, it converts it to wav mono.

What program are you using to extract audio, convert to wav and make mono?

Thank you very much!

Thanks. I guessed so, but I thought maybe someone knows her by name anyway...

https://de.spankbang.com/7auv3/video/oriental+cafe+who+is+the+maid

I can only agree. I am more than happy if I find chinese subs which I can translate with deepl to get at least a rough understanding of what they are talking about, so any language is more than appreciated!

Haha, actually, I have tried that one, but the results weren't particularly better. Some parts got improved and other got worse, so not worth the effort.

Interesting. I've compared the Vosk model with Autosub and it seemed that Autosub did a much better Job than Vosk. May have to do some more testing. I've never tried the Google Vision API but isn't it just for pictures and not for transcription?

Were you using the Big Japanese Model which can be used with Subtitle Edit? Because I've tried that one a week ago and it didn't seem much better than the small one (which was aweful)

Looks pretty good. Do you mind sharing the name of the app you are using?

Hi All I did some work again. A new pack with extracted subs from chinese hardsubs. As always, the files are only cleaned from the most obvious OCR garbage and come from different sources with varying quality. Hope you enjoy :)

Different colors and it scrolls in:

Hi All A new pack with extracted subs from chinese hardsubs. As always, the files are only cleaned from the most obvious OCR garbage and come from different sources with varying quality. Hope you enjoy :)

It seems that the Deepl website got once again changed. Is anybody able to get the new css selector for the button element? Up to now I was always able to find the new one but this time, nothing I tried worked. Any help is much appreciated. Never mind. Figured it out.

Hi All A new pack with extracted subs from chinese/vietnamese hardsubs. As always, the files are only cleaned from the most obvious OCR garbage and come from different sources with varying quality. Hope you enjoy :)

Hi All A new pack with extracted subs from chinese hardsubs. As always, the files are only cleaned from the most obvious OCR garbage. Hope you enjoy :)

Thanks a lot!

Hi All I've found this video 259LUXU-1459, but no name to it. Does anybody know her name or other movies of her?

Did you download the hardsub from supjav and if so, what program did you use? I've tried multiple options, but none seems to work.

Hi All First of all, I have to apologize. It seems that some of my recently uploaded subs had a bad source material because after trying to make sense of the plot, it seemd as if two movie subs were cramped into one. I haven't watched them, only read through the sub file. I've fixed DASD-958...

Hi All I've again done some work and converted new subs from chinese hardsubs and translated them with Deepl. Hope you enjoy :)

Hi All New pack with subs which I converted from chinese hardsubs and translated with Deepl. Hardsubs had different sub and video quality. None of them should be included in previous packs or RUNBKK releases. I only checked for OCR gibberish, so its normal machine translation. Hope you enjoy :)

Does anybody know who she is? https://de.spankbang.com/6j1i2/video/perfect+jap+babe

If you don't know python, then I suggest you watch a youtube tutorial for the installation and basics. Then it should be clear.

Sure, I'm using the tutorial provided by someone on Akiba: https://www.akiba-online.com/threads/tutorial-ocr-chinese-english-movies-hardsub.2005149/. Then just use deepl or google translate for the last step. The results are about 90-100% accurate if you have good source material. Just be aware...