That's interesting. I was struck by how clear the vocals are on
your song with Feng Yi and wondered whether much work on phonemes / voicing etc was necessary achieve that clarity.
I didn't really have to work with the phonemes much to get the pronunciation that I wanted. One of the reasons I chose Feng Yi was because I'd listened to all the voices, and she had one of the less accented voices.
There seem to be more problems with the lower register, so staying in the mid/high register helps. And the Chinese voices seem to handle phonemes like
t and
d, where the coarticulation seems to present issues with Japanese voices.
Honestly, I probably need to work a bit more on getting the vocals to sit inside of the mix, but I like to make sure the vocals are never buried in the mix. So I'll go through the vocals after they're rendered into the DAW, slice up the waveform and normalize them - sometimes from syllable to syllable - and then go back and reduce any unintentionally amplified consonants in the process.
For example, here's the final phrase, and each of the lines is where I've sliced the vocal:
Fortunately for me, Reaper automatically creates fades at slice points, so edits are seamless:
This is a bit tedious, but it only takes a couple of minutes. Plus, it gives me a chance to really listen to the vocal and catch any words that have typos that I somehow missed.
This gives me the ability to control the clarity of the vocal without having to rely on a compressors or de-essers. This track didn't require it, but I'll also use TrackSpacer to carve out space when instruments are encroaching on the vocals.
And, of course, mixing the track so the background stays subordinate to the vocals helps.
