I use the pitch accent script, and put a real focus on mimicking the pitch accent for each vocab. So when I came across Kenichi’s reading of 元 (moto) I was confused because it sounds like he’s pronouncing it mo(H)to(L), whereas Kyoko is pronouncing it mo(L)to(H) as expected (odaka).
From what I can tell from here, when used adverbally it is atamadaka, how Kenichi is saying it. When used as a noun, it is odaka (or heiban apparently), how Kyoko is saying it. Since WaniKani lists it as a noun, I guess she is right.
I have noticed this in other words too. I always check an alternative pitch accent source when this is the case. The Apple Mac Japanese dictionary, Forvo, or http://www.gavo.t.u-tokyo.ac.jp/ojad/eng/phrasing/index are great sources.
It makes me wonder how the voice actors for Kyoko and Kenichi work, especially if pitch accent can change depending on the usage of a word. Do they just read the word however feels natural to them (assuming they’re from Tokyo)?
It would have been super cool if, as part of rerecording all the audio, they took the time to document all of the pitch accent info and make it available via the UI and API natively. It’d be a good sanity check for ensuring the voice recordings were consistent (it’s very confusing when there’s 2 examples and they’re completely different!).