Im a user of SynthesizerV from Germany. I propose an voicebank understandable in all languagues with individual configuration of more powerful consonants (or less powerful).
Ritchy’s pronounciation of consonants is a great improvement for acustic power but he doesn’t give me the full controll over powerful consonants.
This is not practical with modern technology. AI voice databases are based on the voices of real humans, and even with cross-lingual synthesis there is a noticeable accent based on the original voice provider’s native language and accent.
Dreamtonics would need to find a voice provider who is fluent in every language and capable of singing in all of them with a near-native accent, and then they would need to record samples of that person singing in all those languages.
Yes, the IPA exists, and surely there is overlap in languages to reduce the total requirement. But it’s rare enough to find a person who can even produce all of the listed sounds properly (not to mention diphthongs and the like, which are more than a single IPA symbol).
I don’t think it’s a stretch to say there likely isn’t a person alive who can correctly pronounce all the required sounds while singing in a near-fluent manner.