Achieving certain sounds with Japanese voicebanks

When making covers in Japanese with native voicebanks I often still fail to replicate certain pronunciations. This is the case when the singer sings something that sounds like it’s “in between” my possibilities of replication.

Ikimono Gakari - Blue Bird

When she sings “Modoranai to itte” the “te” at the end sounds a lot like a “ta” sound.
When sticking to the original lyric it’s technically correct, but it doesn’t really sound great.
When changing it to “ta” it also isn’t quite correct and it might be perceived as a wrong lyric which I want to avoid.

RADWIMPS - Suzume feat. Toaka

The song starts of with what sounds like something in between “dududu” and “dururu”. Again, both of the options sound off and I’d need something in between to make it perfect.

I have tried being creative and playing around with everything but it feels like this is something I just can’t achieve. I’m not even sure if a voicebank can do this, or only to a certain degree?

I usually switch the note to English or Chinese and “borrow” a vowel phoneme from one of the other languages (and I also add a dictionary entry so I can easily enter and adjust the same phoneme sequence elsewhere in the project, if needed).


You can also try to cycle the alternate phonemes or adjust the phoneme strength.

In all cases it’ll vary between voice databases just how much flexibility they have in their pronunciation, especially when the base language (Japanese) has such a small set of vowels to begin with.

In rare cases I choose a completely different voice database that sounds similar enough, then use that to fill in the one missing phoneme and either overlap them or quickly cross-fade between them in my DAW (I had to do this once to get Maki to draw out an English “f” sound, for example).

1 Like