Reviewing Language Reactor, Migaku, jidoujisho, etc. as an alternative to Subs2SRS

[emk]

Anyway IMO we should keep this thread on the topic of tools for working with video, not just text.

Maybe my post was off topic; I'm not sure. I find it a bit confusing, since there's such overlap in the tools' functions. Language Reactor allows the user to import a web page, for example, for LWT-like reading. Lute will associate a sound file with a text, for listening while reading, but it won't make audio flashcards. The headline feature for Subs2srs is making flashcards from video, but it also can make flashcards from audio. Etc.

Yeah, there are a lot of tools out there which do LWT-style reading. Almost too many to keep track of. So I was thinking it might be worth limiting this particular thread to tools which:

1. Support working with video and/or audio, and
2. Claim to support making audio flashcards in some fashion.

Of course, if a tool also supports LWT-style reading in addition to audio & video, then that's worth mentioning in the review. Does that make sense?

I also tried out asbplayer's in-browser video player yesterday. I liked the idea of it, but found the usability poor because it just didn't seem to support any of the usual keyboard shortcuts that we're used to with video players to skip back and forward etc. Still, it could be a good option to have if you want to play local files but benefit from asbplayer's flashcard creation and having subtitles rendered as in-browser text that can be handled with extensions like Yomitan.

I'm disappointed that asbplayer's video player and parallel subtitle handling sound so weak. Those are both really easy things to fix, and asbplayer sounds like it's otherwise excellent.

I really appreciate the effort everyone is putting into investigating these tools. Audio cards are promising, but the tools for working with them are still pretty unfriendly and/or poorly known.

[loopernow]

This is neat! I ran across your web pages about Subs2SRS a few weeks ago! Was going to try it and then, like you said, realized the geekiness factor was more intense than I was willing to invest!

I'm enjoying your reviews here.

Someone mentioned lingopie earlier (end of page 1?). Do you think you'll try it?

If not, maybe I can post a review here later. I just signed up for the 7-day trial a half-hour ago…

--

Edit: LingoPie does not support flashcard export!

[Carl]

Yeah, there are a lot of tools out there which do LWT-style reading. Almost too many to keep track of. So I was thinking it might be worth limiting this particular thread to tools which:

1. Support working with video and/or audio, and
2. Claim to support making audio flashcards in some fashion.

Of course, if a tool also supports LWT-style reading in addition to audio & video, then that's worth mentioning in the review. Does that make sense?

I was hoping you'd weigh in on defining the topic of this thread, emk. Thanks for doing so; that's clear and it makes sense.

Someone mentioned lingopie earlier (end of page 1?). Do you think you'll try it?

If not, maybe I can post a review here later. I just signed up for the 7-day trial a half-hour ago…

Here's a tip for the Lingopie free trial: I recently signed up for the free trial, and after 6 days decided that the selection of German-language materials was too small on Lingopie, and that I'd look again at Language Reactor. When I canceled, they asked me a question about why, and the answer I picked (I forget what it was) led to an offer to extend the free trial by two weeks. So I accepted! I've used it to watch a couple episodes of WaPo Berlin on Netflix, a cheesy police procedural about a newly created police unit based on a boat on Wannsee. Both episodes begin with someone boating on Wannsee and then suddenly running into a floating corpse. I'm curious to see if they have other plot ideas...

OK, that last bit clearly was OT. Sorry!

[loopernow]

I do want emphasize some key differences between the projects...[E]verything related to audio cards [with Language Reactor] is fatally flawed, in my opinion, by the use of synthesized audio voices.

So I was really intrigued by this thread (thank you @emk!) and tried Language Reactor and Migaku. I found that, actually, when creating audio cards from Netflix subtitles, Language Reactor uses original audio!* And image capture. So I'm thinking the limitation to non-original audio may be specific to YouTube with Language Reactor (I don't know for sure; I only tried Netflix).

* - Edit: I found that it depends on the source. For example, Money Heist exports original audio. But an anime, Blue-Eyed Samurai, exported machine-generated audio. Perhaps it has to do with licensing, I don't know.

Migaku, on the other hand, just feels clunky, in many little ways. But if you're using it to make audio cards, it has lots of advantages[.]

Most of the advantages emk listed are also supported by Language Reactor, at least in Netflix.

Migaku is definitely more clunky-feeling, but these tools are mostly the same in terms of capabilities, with Migaku being slightly more powerful. They both have hotkeys for repeating subtitles, jumping between subtitles, etc. They both can export subtitles to Anki with original audio and screenshots in Netflix. They both can automatically pause video at the end of a subtitle, and hide subtitles until the user presses a key.

They're also priced similarly.

Migaku is much, much more granular. You have all kinds of different options for the format of the exported Anki cards, among other things.

Two points:

• Migaku can bulk export all of the subtitles in an episode at once to Anki.
• Migaku has two different Chrome extensions. The "legacy" extension is the more capable extension. I did not try the non-legacy extension.

I think @amk also mentioned at some point that the subtitles in Language Reactor did not match the audio. This appears to be a function of each individual video, not of Language Reactor. I watched a bit of Neon Genesis Evengelion and a bit of Casa De Papel/Money Heist in Netflix, and the Money Heist subtitles are pretty much word for word matching the audio. Whereas NGE's subtitles are absolutely not. I think this is just what subtitles are available from Netflix for these shows; the extension being used doesn't matter (unless an extension loads its own subtitles, but I don't think either of these tools does that).

Language Reactor actually has a "machine language translation" feature for translated subtitles, but this doesn't resolve the underlying issue of whether or not the original language subtitles are themselves faithful transcriptions of the audio. What is machine translated is the original subtitles, not the audio. :/

So, to me--these are pretty similar if used with Netflix, if you use the legacy extension from Migaku. Migaku takes much more setup to get started, as mentioned by emk. Migaku is a bit more powerful if you are trying to export an entire episode of flashcards in one go, or if you have very particular goals for what you want your exported cards to look like.

That is my clarification of these two tools after a little bit of fiddling around with them!