-
Notifications
You must be signed in to change notification settings - Fork 71
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Subtitle Edit 4.0.9 and Purfview's Faster Whisper XXL Still Broken Merging Subtitle Lines #339
Comments
Try to run directly from faster-whisper-xxl.exe using command. Purfview already released a new version with .bat included (drop files in bat) & check if the result is the same as with Subtitle Edit. |
Those SE screenshots doesn't tell me anything. Not interested in those videos or whatever you posted there in SE issues too.
|
Open a new issue when you can provide all those things described in the above post. |
Thank you as always for the quick reply, Purfview, I have here one new example, this is with the standalone Faster-Whisper-XXL r194.2 (before today's update with r194.4), not using Subtitle Edit at all. The .json file is included now. The commands were the default (I only changed it to the large-v2 and to Translate to English), or this from the .bat file: faster-whisper-xxl.exe %file_list% -pp -o source --batch_recursive --check_files --standard -f json srt -m large-v2 --task translate This is a 1-hour video from the DdeunDdeun channel on Youtube, I've also included the official English and Korean subtitles as comparison. Pinggyego/Just An Excuse Episode 58: https://www.youtube.com/watch?v=B4bOGORy58E The problem (with Korean, Spanish, probably a lot of the other languages as well, especially when the video or audio is over 1 hour long) seems to be that the lines are merged because of the --standard command or something along those lines, and so the final output with the .srt file in English has the random capitalized letters, missing periods and commas, etc. I apologize again if it's not reproducible, I'm using the latest version (not the r194.4 that was uploaded a few hours ago) of Faster-Whisper-XXL, and even with regular Faster-Whisper back then, it definitely looks like the situation is due to the introduction of the --standard command being default or something like that. I can't really describe it further in detail as I'm still pretty new to all of this, but it seems to be related to the updates around the time period between Subtitle Edit 4.0.3 and Subtitle Edit 4.0.4. The previous default (for Subtitle Edit) didn't have as many breaks or " Thank you again for the fast response. The previous examples from several months ago also exhibit the same behaviors (they were done through Subtitle Edit instead of standalone Faster-Whisper(-XXL) though). github subtitle edit broken translation for korean chinese etc.zip github subtitle edit new broken translations for korean chinese spanish etc.zip |
"--standard" preset activates "--sentence" and sentence breaks/joins lines, if punctuation is missing then there will be wrong joins.
Subtitle writers doesn't change capitalization nor adds nor removes anything in transcription, it can only break and join lines according to the settings. In Read the help to know what some setting do. |
Hello, thank you for the newest updates.
With the newest Subtitle Edit 4.0.9 beta update (and every release after the final version of Subtitle Edit 4.0.3, or after December 23, 2023) there's still the broken/merged/random capitalization/no commas or periods/etc. issues that I described here: SubtitleEdit/subtitleedit#8634 and SubtitleEdit/subtitleedit#8209
I've also posted it again on the Subtitle Edit page since the default --standard option (can't be toggled on or off right now, like the settings are not saved) in Subtitle Edit is what's creating the issues: SubtitleEdit/subtitleedit#9035
Once more, thank you so much Purfview, really appreciate all the work and help you've done. I apologize again if I was unclear or rambling, the specific examples (with videos and images) are in the previous comments.
The text was updated successfully, but these errors were encountered: