Cheers for the app @christianwu ! I've given it a go with some Japanese recordings I already had and it works pretty well, though it does throw in a few random English words that weren't actually said. Not a massive deal for me though.
Just a quick feature request: at the moment, if I switch between the transcription options (like moving from English to the CN+JP+KR pack), it seems to re-download the model every time. I often jump between Japanese, English, German, and one other European language, so the constant re-downloading is a bit of a pain. And I am often outside WiFi. Would it be possible to keep the models cached on the device so they don't have to be downloaded every time I switch? Or perhaps add an option where you can just pick which language to use for a specific recording?