Added the ability to save LLM settings

• Now you can save LLM's base url and other related settings.
• Fixed little bugs here and there
• Now you can convert a portrait video into a landscape video, or vice versa with hard-coded subtitles.

Finally a good release

This release has many changes. Seems it worth the 2.0 version.

• Google translation is now working
• Many bugs are fixed. (Once again, but there should be bugs left behind.)
• The package is different. You need to:
  1. Install Python 3.10 or above.
  2. Run the "pre_install.bat" once. A folder named "venv" should be created.
  3. Now start using it with "run.bat". No need to run "pre_install.bat" any more.
     Enjoy !!

Update: Oops ! The "pre-install.bat" and "run.bat" were wrong. Please download it again.

Many fixes

Now you can see the program log in "Transcribe / App Log" tab, which make it easier to see what's actually going on in the background.

• The Google Translate is not working yet. Don't use it.
• There is no more built-in free API. If you need AI feature, set the Base URL and API Key.
• Word splitting and smart sentence breaking should work now, but it requires LLM settings.
• It's a bad idea to use "Word Splitting" and "Translate to English" at the same time in FasterWhisper.

More control in video synthesis

• Now you can make portrait translated subtitled videos from landscape videos.
• You can either choose a picture background, or default to the video's zoomed fuzzing background.
• The ASS style is working great !!
• You can control subtitle's vertical offset. The zoom size of video and/or subtitle. Very useful for portrait videos.
• It seems the AI single translation is working fine. It takes less time and effort compared to the optimize+translate method.
• Now the batch tasks display a lot more information about each task, with some portrait / landscape control.

bug fixes and new package

This is a package using python 3.13 and Visual Studio virtual environment.
For reasons unknown to me, Kaka's original package won't work with new code anymore, and I have no idea why.
It's a very different package, so its size is bigger than ever.
Just extract it then run the "run.bat" file.

• Fix some subtitles generating issues.
• Make "AI single sentence translate" better.
• Now runs in Python 3.13
• More multilingual fixes.
• Add subtitle time offset setting. Simply add or subtract milliseconds to all the subtitle lines.

Another one with many fixes

Most fixes are internal.
The obvious one is the added translation choice.
It will have 3 types:

1. Optimize and translate. It will perform AI optimizing then translating.
2. AI single sentence translation. It will feed 10 lines at a time to AI and get the result back.
3. Google translation. This is not implemented yet.

Various updates and fixes

• Allow 2 tasks run at the same time, so that one does audio transcoding while the other does transcribing.
• If the video's audio codec is "aac", "mp3" or "pcm", the program will skip the transcoding and do directly copy. It greatly accelerates the overall speed. FasterWhisper itself will transcode all audio to wav file itself anyway. No point in doing slow transcoding.
• Batch process cards looks better with wider size and more info.
• Add a language widget in the right-lower corner of the home page, so now changing the language is super-easy.
• Now if in the setting "need video" is disabled, the Batch Process tasks will default to be "Audio/Video transcribe".
• Other program bugs.

Add "minimum sentence time length" feature

A lot of times some sentences in subtitles are just too fast to read.
Faster Whisper XXL creates a lot of this issues.
You can use SubtitleEdit to fix it, but I find it exhausting to fix the subtitle files one by one.
So now in the settings there are 2 new entries. If you enable this feature, by the end of transcribe this program will go through the ASR_DATA and fix sentences that have too little time length. Default value is 1500ms, which is 1.5 seconds. You can set the value yourself.

Translation is finally done properly

Well, I finally learned how to do the translation with QT. Here is the result.
If you want to update, you just need to overwrite the "app" and "resource" folder.

Updated package.

• Translated most to English.
• In Options now you can choose output subtitle format, file name prefix and suffix.
• Now when using FasterWhisper "Translate to English" feature, it will use max line width of English instead of Chinese.
• Add Exit, Shutdown, or Suspend when batch jobs are done ( experimental ).

To update your existing VideoCaptioner, just copy the "app" folder, otherwise extract all files and run it from there.