Enable import of srt-file over existing lesson
under review
A
A.J.
I often import from youtube, if it has subtitles in Finnish, and most often the sentences are all broken up into very same lines that fit on a screen, but are not full sentences. Translation will always be a bit off, as will the timing of speech (because the sentence is still ongoing, pauses between words can be very short). Also, working in sentence mode does not give me a full overview of what the speaker means, because it is a fraction of a sentence.
So, my proposal is to allow an srt-file (that I generated myself) to be imported over the official text. Advantages are:
- I can create full sentences, if I want to and get time-stamps that fit sentences.
- I get translations per sentence in stead of per fraction of a sentence.
- I enjoy all the defaults, that Lingq created when importing from youtube, except for the text and time-stamps itself.
- I no longer have to import an srt file as a book, where I have no control over the other lesson parameters like course.
I don't see any downsides, as overwriting with my own srt file, is optional, not mandatory. So, anyone who doesn't want this overwriting, doesn't have to do it.
H
HsingH
This would be good, as there are better and faster tools for doing timestamp on desktop manually like Aegisub.
Mark Kaufmann
A.J. How do you generate the new srt file? Do you take the automatic transcript and then manually adjust it/add timestamps?
A
A.J.
Mark Kaufmann Thanks for considering this request, I appreciate it! Do I change time stamps? No, I use a local whisper variant that can transcribe text. This program has the option "-sentence" causing it too create full sentences, unless they are really too large. It is not perfect, so it might make mistakes, but all in all it is pretty reliable.
Mark Kaufmann
A.J. Do you rip the audio from the YT file first for whisper?
A
A.J.
Mark Kaufmann If I have to. What I want to do, is just import through the LINGQ import addon and then overwrite the lesson from the subtitles with my own SRT file. That way, I have sentences, if I want to. The audio will still work as with the YT import, just different time stamps.
Mark Kaufmann
Also, can you not simply import the srt file using the file importer and attach the YT link to that lesson after it has been imported?
Also, we are working on improving our import from YT which now provides better auto generated captions for most videos so this issue should disappear over time.
Mark Kaufmann
A.J. In the short term you should be able to create a new import with the SRT file and then attach the YouTube URL to that lesson to achieve what you're after, I believe. Let me know if you have any issues with that.
A
A.J.
Mark Kaufmann The thing with importing srt is that the import utility does not accept any parameters. It creates a new course with the same name as the lesson, because it was created to import books. That is very inconvenient as I have to change everything: move the lesson to the course I want it to be in, change the lesson name if necessary, add image etc. Using the standard youtube import sets all attributes correctly and I can determine where it goes. Maybe if you would create a non-book import where I determine all the parameters (title, course, difficulty etc) it would be tolerable. Also, even if the import from YT is improved, Do I get a say in whether it breaks on sentence or on subtitles?
Mark Kaufmann
A.J. Those are good points and are things we will try to address in the new import page. Most Youtube videos now seem to have auto generated subtitles with punctuation so will no longer be imported in chunks at all. I would not expect for there to be a way to choose chunks vs sentence in those cases. We are not currently looking to reformat imports with the old format of chunked subtitles as we tried that previously and it created as many problems as it solved it seems. Anyway, this is good feedback and we will keep it in mind as we improve importing going forward.
Expect to see a new import page with better options before the end of the year. And, from there further improvements over Q1.
Mark Kaufmann
marked this post as
under review