I would pay 50 USD extra per month if LR had this function

Request
1

I would pay 50 USD extra per month if LR had this function:
Pretty much all Japanese media has subtitles embedded into the video itself.
The auto subtitles for Japanese is pretty bad.
If LR had a simple optical recognition function (even the free open source ones would do) then the amount of Japanese media that could easily be used with LR would go up by orders of magnitude just with the Youtube Extension alone.

If you make the function in a way where the user has to download the full open source OCR on his own computer, then that would remove all network risks and increased network demand on your end. And you could make it an optional purchase to not deter other customers. Developing this function should be very cheap and easy to do, since all you will be developing is a connecting program between the OCR and the already built LR.

I would be very, very happy to pay extra for this function.

(If the developers of LR are reading this: I am very grateful for your program. Thank you for making it.)

1 Like
2

its the same for mandarin. proper subtitles whether through optical recognition would be great on youtube
or proper speech recognition

3

bro use capcut to extract japanese sub. it is 100percent accurate

1 Like
4

i think you need to pay for the function if i am not wrong, i havent used capcut in a long time but some options maybe payed

5

there is capcut as someone in the comments mentioned, i must say that the quality of the subtitles must be really clear for OCR to extract the text

im not studying japanese or chinese but personally i use shareX to extract text for english and french, it comes with OCR but its pretty bad so i just use it to take pictures and shareX copies it to my clipboard then i just pop it into google lens so google lens can extract the text. that is another option if you would like to check it. i must also say that maybe it can be quite confusing trying to set the “copy image to clipboard” configuration but its free. i think OCR will always have issues when it comes to extracting text so my recomendation is to find many tools as you can with the same function if an issue comes up

6

CapCut subtitle generation is more accurate than YouTube’s automatic subtitles, but by default, it creates short lines for subtitles.

It’s possible to merge these short lines for better subtitles in CapCut. It’s not very difficult, but it does take time.

Also CapCut does not allow exporting subtitles by itself but there are workarounds for this.

Even so, it would be very useful to allow adding manual subtitles for translation in LR. Users can create better subtitles using CapCut or another software. If LR allows the use of external subtitles, this can be a viable option.

7

Edit: It is possible to do this yourself manually using many programs and clicking a lot and copy and pasting. But the point is to make learning languages more effective.

My idea of the program I want is this:
I click a youtube video or play a video game. Language reactor automatically detects subtitles on the video itself and displays those in the LR subtitle track. And I can just point and click on the words I don’t understand.

This scenario is not technologically impossible and is also not expensive to make. LR already has the auto-subtitle machine learning option on Netflix.

1 Like
8

It is possible to add custom SRT files to YouTube, etc., using an add-on called ‘AsbPlayer.’ Unfortunately, it is not possible to use these custom subtitles with Language Reactor. It would be very useful if Language Reactor supported custom SRT files for YouTube. It already supports it on languagereactor.com’s media > custom player page, so I don’t think it’s a technical issue.

If this was possible, we could use Capcut-generated subtitles with YouTube, which would be very helpful.

Of course it would be wonderful if language reactor integrates its netflix ML to youtube but this seems a large task for youtube’s content. May be a request based system can be implemented.

9

i didnt know about capcut. thank you so much this helped me a lot!

10

I couldn’t agree more with your point—burned-in hardsubs are by far the biggest barrier when trying to immerse in Japanese YouTube videos and web media.

Since LR hasn’t implemented this, I actually started developing a browser extension called YuzuLingo to solve this exact problem.

It automatically reads hardsubs off the video stream in real-time using local OCR and turns them into fully selectable text. That way, you can hover over words for instant dictionary lookups (via Yomitan) and mine sentences directly into Anki without typing anything out.

And just like you suggested, the OCR engine runs 100% on-device on your own computer, meaning zero network lag, no privacy concerns, and no extra server costs. I’m building it specifically so we don’t have to miss out on the huge amount of hardsubbed Japanese content out there!