It's pretty good straight out of the gate. I think giving customization of fonts is an excellent idea. There's huge variation in this area and font-switching is a definite stumbling block for anyone doing JSL. If someone has gotten used to a sans serif font from a textbook or Anki deck, for example, the more visually complex serif fronts used on official documents and exam papers require extra mental effort to parse. Likewise being able to switch color combinations easily is a good idea.

Might be good to allow the kanji/vocabulary to be filtered by JLPT or Jouyou stage. Picking multiple sets on the kanji units was a bit tedious, it's be nice to 'pick all' for a drill (but I was using it to test myself rather than learn new ones). I don't understand the pick options (pick, reverse, input, output) - they seem superfluous and perhaps need tooltips. Maybe add audio recordings at some point, although that's a bunch of work. You can use AI to generate it of course and it will be mostly correct as far as individual words go, but Japanese AI voices still seem to get pitch shapes and timing wrong sometimes.