Hello!
It’s been a month since my last update regarding voicebank development. Back then I was simply testing the voicebank’s compatibility through various Indonesian dialect. As for today I want to share you my journey on conquering my fears and anxiety to ask for help to finally make my dream comes true.
It all begun when Cadlaxa posted about releasing his OpenUTAU Dictionary Editor (check it out here: Cadlaxa’s post of OpenUTAU Dictionary Editor). At first I do not know how it works, I don’t even know how to install it. But through patience Cadlaxa helped me to be able to install it. (Turns out the problem was that I don’t know that I have to add path for Python to work……….). And there I go on a journey of making CMUdict for my Indonesian CVVC SupaMegaLite reclist/voicebank.
As I slowly and manually works my way through adding words to it, I realized a problem.
“For it to be useful, it needs a vessel.”
I realized that my dictionary would not work unless there is some kind of a program that reads and process it. And that is when I have an Idea of looking through the French dictionary as references. And as I look through it, I was intrigued by how it was written. I wanted to know more and I gather my courage to ask a friend of mine, Mim (the developer and voicer of UTAU/Diffsinger Mimosa) about a way where I can use my dictionary.
At first Mim asked me about the logic of my reclist/voicebank. And after explanation and conclusion, we reached an agreement that my ID CVVC SupaMegaLite reclist/voicebank works with this logic:
If it's a starting V -> (eg: aku [- a] [a k] [ku]) If it's a starting CV -> (eg: bapak [- ba] [a p] [pa] [ak]) If it's a starting CCV -> (eg: gravitasi [- gra] [a v] [vi] [i t] [ta] [a s] [si]) If it's a starting C -> (eg: strategis [s] [tra] [a t] [te] [e g] [gi] [is])
If it's a VV transition -> (eg: kelamaan [- k@] [@ l] [la] [a m] [ma] [a a] [an]) If it's a CV transition -> (eg: apanya [- a] [a p] [pa] [a ny] [nya]) If it's a CCV transition -> (eg: kalibrasi [- ka] [a l] [li] [i b] [bra] [a s] [si]) If it's a V C transition -> (eg: kata [- ka] [a t] [ta]) If it's a VC trasition -> (eg: bergumam [- b@] [@r] [gu] [u m] [ma] [am]) If it's a VCC transition -> (eg: mengagungkan [-m@] [@ ng] [nga] [a g] [gu] [ung] [ka] [an])
If it's an ending VC -> (eg: main [- ma] [a i] [in]) If it's an ending VCC -> (eg: paling [- pa] [a l] [li] [ing])
And off Mim go, working on the phonemizer as I was busy researching word banks and dictionary I can lend to feed to the CMUdict. After 3 hours of Mim’s godspeed coding, they manage to build a vessel for the dictionary, the ID CVVC SupaMegaLite phonemizer. Here’s a preview of Mim’s hard work making my logic works!
ust by zimtroeschen/zimty, you can check their version here: UTAU DEMO
And there it is, thats how my quest of making an ID CMUdict that ends up with Mim’s ID SupaMegaLite phonemizer starts. And I would probably will continue for the next several weeks as another challenger has come to challenge me and that is: