Sorry I didn't jump in here earlier: it actually sounds OK to me, it's a bit low at the beginning but if you compare her chang2 to her chang3 there's a pretty clear difference. But we could certainly tweak it - maybe clip off a bit at the beginning so you don't have as much of a low section.
To be honest, we've never been entirely happy with the female recordings in general and have been thinking for several years now that we ought to commission someone to do a brand new set - would also let us cover some vocabulary we haven't covered before, and would let us record an alternate set of shorter syllables (as we've been doing for our new Cantonese recordings) to stitch together into words.
We've been hesitant to do that mainly because we're not quite sure how interested people are still going to be in pre-recorded audio now that we have our new TTS add-on - it's female only (and the library we licensed doesn't support installing two voices in the same app), so it still leaves a definite need for a male voice at least, but if the sentiment among our users is that they don't really care about female recordings anymore once they have TTS then we may let the issue drop, or at least limit our new recordings to single syllables only. (it's much cheaper / easier to commission ~2000 recordings for single syllables than ~30,000 for multi-syllable words, and while that comment might lead to accusations of stinginess, it's really more a matter of budget allocation - do we spend the money on hiring a contractor for that, or hiring one for something else like improving our new character decomposition and thesaurus databases)