Getting note duration working required a whole lot of extra information:notes, syllables, pitch, segments, and so on.
The first two lines of "Twinkle, Twinkle, Little Star"
It still sounds pretty robotic, but very clear.
I've implemented interpolation across frames so the output is smooth, but can't hear any difference.
I haven't implemented cross-fades when joining disparate source segments - something else I should look into.
The next obvious thing to look at is an English to Phoneme routine. I should probably also port over the pitch line I've written, to see how much (if any) it improves the output.
I also need to have a closer look at the reclist to see what's missing, and how to automatically handle missing phonemes.
No comments:
Post a Comment