I don’t use this option very often but I redcorded an anime and uploaded it. After doing so I saw that YouTube is not allowing people to export the subtitles making this tech/option a bit more relevant. Anyway, it seems to have skipped transcribing the first several minutes of the audio. Is this common? Should I just delete and reupload it again? I have a limited amount of transcription per month so I’m not so keen on reuploading stuff over and over.
Can you try to re-convert that file to mp3 by using this online tool:
and then give it another try with importing?
2 Likes
It worked. So I should run all my sound files through that site before importing?
I’m routinely having the same issue. Some pieces are skipped, others are duplicated and put in random places. The most weird effect was when the video was transcribed up to some phrase, which was then repeated many dozens times. And I also noticed that this problem depends on the audio quality, but I haven’t find an optimal solution so far. I’ve tried the recommended converter (with default settings) and it actually made a worse result for the current episode I want to go through than my last attempt. Maybe one can add some recommendations on the specific audio quality to be used?
Also the issue almost never occurs in podcasts. It is only movies with all kind of sound effects, gaps between conversations, etc.
1 Like
@kraemder Thanks for letting me know. Yes, sounded to me like that was a coding issue with your file, so running it through a converter tool usually fix it properly.
@pchuhaseraya Was your file in mp3 format too?
1 Like
I had exactly the same issue last week after uploading an mp3 audio-file for transcription (endless repetition of the same sentence). I deleted the upload and imported the same file again. But this time the transcript contained a lot of other mistakes (such as skipping whole sentences).
Making and uploading a completely new audio-file from the same source (youtube-video) finally solved the problem.
The audio contained just one clear voice, no music, no background noise.
Yes, they are mp3. It seems to me that too high quality almost always gives the worst issue with transcription, while something average (120-190 kbps) is better. But to be honest I haven’t tested that too thoroughly. With the latest uploaded lesson, my “original” mp3 (saved with Audacity after cutting out some extra stuff) and the converted with the link you’ve provided were missing exactly the same piece about 1:30 min long (the piece itself, when I cut it and uploaded separately was transcribed without issues). But the converted file was missing another larger piece before that as well. If you have access to my private lessons, I can provide links to both (though I’ve already edited one of the transcriptions). Or I can upload both files somewhere, if needed.
@pchuhaseraya Sure, feel free to send lesson URLs to support@lingq.com and I will take a look.
1 Like