Order of an entry's different pitch accents and the generated audio

zxcv

I was wondering if the audio generated for a word with multiple pitch accents corresponds to the pitch accent that is listed at first. For example, for 占い
it would be pronounced as heiban in the audio instead of naka? If so, then we could also assume that the heiban pronunciation is the most common for that word?

fr3nd

The audio and the pitch accent, both come from two different unrelated sources. While the pitch accents are ordered by frequency, the generated audio is supposed to be the most frequent one, but at the movement I'm afraid there is no way to make sure what pitch accent the audio is reproducing.

I have on my to-do list a task to improve the generated audio quality. I will investigate if it's possible to generate a different audio for each pitch accent pronunciation.