My thoughts pretty much line up with this. She's not incomprehensible, but the pacing/construction of the sentences definitely sounds like a Japanese speaker learning English's expectation for a sentence rather than a native English speaker's expectation of the same sentence. We'll see what the clarity's like once the product is in the hands of proficient English-speaking users. Might be the same. Might not be.In regards to her English talk bank, if I remember correctly, doesn't Cevio talk allow you to adjust the pitch/timing/etc of the different parts of the sentence? Because to me, those are the only things about these demonstrations that sounded like they could use some work. her pronunciation of the phonemes themselves didn't seem too terribly accented. If the software works the way I think it does, it may not make a good plug-and-play TTS, but it will be great for generating female character voices for animations and talkloid type things.
I'm more excited to see the English SynthV bank now.