As a Japanese learner, I was thinking about doing this. I didn't though, here's a couple reasons why.
- For audio memo, each target would be a mora, which is less compact than combining 2 letters to make a syllable.
- My Japanese vocabulary is bad, meaning kana pair images would be impossible to reliably come up with.
- 26 letters maps better to the 23/24 letters you need than 40something kana. I imagine a kana scheme would involve each face being a different consonant, with the four different targets on the face being different vowels. Since Japanese has 5 vowels, this is really ugly IMO.
Saying this, I'm really curious to how BLD solvers in Japan memorize. My Japanese skills are still definitely too weak to understand any videos, and reading through a written guide would be painful. Maybe someone could enlighten me?
Hopefully you take my thoughts into account. Good luck coming up with something that works for you!