Until somebody fixes the specifications area, I guess that here would be the place to share the missing documents. I've attached it.
This issue has been always broken in the specifications. First, the original SMF spec only allowed ASCII so every language other than English and Latin was excluded. Then, the RP26 followed the trend, regulating languages instead of encodings. The result was that nobody cares about it.
Of course, there are SMF/Karaoke files out there using every possible text encoding you may imagine, and players need to figure out which encoding was used or ask to the user. For western languages using Latin alphabet variants, the most practical is UTF-8. But for eastern languages that is not the case.
Maybe you can try your files using this open source program. It tries to decode lyrics from SMF files guessing the encoding with the help of the
uchardet library:
https://dmidiplayer.sourceforge.io/