Hi, I rarely do development diary but I need to change that. I should be more strategic with how I approach my projects.

In this (my first development diary) I will be sharing about my new experiments and findings developing an UTAU voicebank. This time is not my normal cup of tea Japanese voicebank, but Indonesian CVVC voicebank.

Illust. by hitunanishi/August

Why Indonesian CVVC voicebank? Because I want to preserve the language in a media I have heavily brain rotted. I figure if I develop this, this will have a lot of potentials and a lot of weaknesses. I want it to be something that can inspire other people to protect and make flourish, this voicebank format.

What are the potentials? Despite it being monogenous language, I believe of combability is maxed we can achieve coverage or what I called cross dialect system between dialects and accent of the many different language of Nusantara. This voicebank, as per my finding can sing in Bahasa Banjaran, Bahasa Minang, Basa Sunda, and Boso Jowo beside Bahasa Indonesia.

What about the weaknesses? Let me guide you through each and every difficulties I’ve gone through as I was using the Indonesian CVVC.

• It has a very floaty odd accent, presumably because I am used to recording Japanese. This voicebank requires you to be more grounded with your pronunciation.
• Vowels are not complete, has less coverage.
• Often times with the engine and how the software works, accent or dialect doesn’t translate well.
• Steep learning curve but once you get around it, it will be a breeze.
• Can be automated but certain phonemes/symbols halts it (looking for a work around).
• So far works with original UTAU but I’m not sure how it is in OpenUTAU.

These weaknesses is there to be my sun, but I will keep flying onto the sun to test the limitation and workarounds. Weakness is there to challenge me, not to stop me afterall.