Watching Frieren season two and just paid for pro… but realised some conversation lines aer skipped with no subtiltes nor translations showns. Any fixes? thanks.
1 Like
It’s likely a problem with whatever speech recognition model is used under the hood to transcribe, not really with LR itself. It’s a common problem with some models, but some are better than others (OpenAI Whisper large models, Parakeet). If it’s a Youtube video then it’s likely coming from Youtube’s subtitles anyway, which I’ve found are pretty bad. Sucks but I personally recommend with any software akin to LR (LingQ being another example) to just download audio/video files and make your own SRT files with a more advanced model; LR now has the media file feature which works well for this workflow. As these models continue to improve we can hope the improvements will become commonplace everywhere.
1 Like
Hello @Ian1,
Thanks for your message!
Language Reactor subtitles on YouTube are based on the subtitles provided by YouTube. If the YouTube subtitles are not properly synchronized with the video (for example, if some subtitles are missing or if they appear earlier or later than the spoken audio), the Language Reactor subtitles may show the same behavior.
We will look into possible ways to improve this in the future.
In the meantime, you can avoid this issue by using the MediaFile section. You can simply download the video from YouTube and upload it there to watch it with properly synchronized subtitles.