The Essential UTAU Toolkit - Romaji/Hiragana UST and voicebank conversion/aliasing

Download~

Yep, it's the all-amazing converter for romaji to hiragana and hiragana to romaji USTs! (Okay, technically it's for romaji to hiragana. For hiragana to romaji, I provided instructions, though.)

Finally, all of our nightmares are over.

I've tested it, too, and it works!

If any of you who know Java or know what you're doing, if you have modifications or improvements, feel free to submit.

I based my code off a text replacement Java source code floating around~

/EDIT: v2 released. v2 now includes support for hiragana to romaji. Also more user-friendly, so it involves running a bat file instead of Command Prompt. I have been informed that it works in a computer with non-Japanese locale, as well. Download here.

/EDIT: v2.1 released. This version is more user-friendly, as it no longer requires conversion to .txt. Also, through testing I have observed that the location of the files is not important, so I have updated the readme accordingly. Download here.

/EDIT: v2.2 released. More accurate readme. Download here.

/EDIT: v2.3 released. Full OREMO reclist support along with capital-letter support. Download here.

/EDIT: This is something you HAVE to read.

I have created a whole new infrastructure with this program and have thus renamed it "The Essential UTAU Toolkit" (but codenamed "Converter 3.0"). This program now supports:
* oto.ini aliasing for adding aliases to voicebanks (hiragana or romaji)
* adding aliases to diphone voicebanks
* converting diphone symbols to a form usable when it sometimes doesn't work on non-localed computers
* removing diphone marks from USTs
* ...and of course the original purpose, converting USTs back and forth from romaji and hiragana.

All the tools a non-locale UTAU user will ever need (and useful things for localed UTAU users too).

You can download the beauty here.

Oh, oh! This is great!

I was thinking somebody could make something of an UTAU plugin for this... but this is really a lot better idea as UTAU can be trippy with plugins sometimes!


Hmm, but to use it is a bit hard for noob level. Why not to add a .bat to make it automatically? Named like "Run me to convert" (or something like that) with a command of only one line:

java TextRep convert.txt mystrings.txt

And ask to double click it in the basic instructions (explain the full method for advanced users).
Hmm, maybe a problem with new Windows? It works for my XP but I have no idea how Vista and 7 uses .bat.


It works really well even in not Japanese locale (but you need to go to Japanese locale to see the letters correctly and not the mojibake). Quite a surprise, that's quite cool if you only use UTAU with AppLocale (I found you can sometimes use kana voicebanks in AppLocaled UTAU, but it's very... hmm, "unstable").


The backup copy is a great idea. You should mention it in the instructions for completeness.


Aside is the hiragana to romaji hint.
It works great (it's going to help me a lot!), but only works if you can open the .txt-fied .ust in Japanese locale (www.romaji.org can't traslate mojibake) so it's not going to help the "only using English locale" users that is probably the main users that needs this option.

All right, I'll add a note about that.

I know nothing about .bat files >< I only know Java...but I'll make one and see how it works out. I'll probably release new versions as I go on.

I can easily make one for hiragana->romaji, too, with the same method for romaji->hiragana. But I didn't expect the other way around to work in non-Japanese locale, so that was a surprise.

Nice work.

I took a quick look at the java. It seems to be working on a byte-by-byte level, which is how I made my TestUST program.

This program you setup will work exactly the same on a computer that doesnt even have japanese installed at all because it is not processing the data as text.

In Spanish locale I can convert a hiragana ust to romaji one. It's so cool, agh. So useful!
I want to note here that I have Japanese compability installed so I'm not sure if with a 0% Japanese computer it's going to work (probably will do for what BlueboyX says).

I was playing around and found you don't even need to rename the .ust to .txt.
If the .bat has the command
"java TextRep convert.ust -otherfile-.txt"
it directly works on the ust. No need of txt thing, heh.
It's normal as .ust is just plain text, anyway.

...okay, I'm going to update that. ^^ I feel like a software distributor now.
Thanks for all of the input, by the way, you're really helping out.

By the way, do you want me to support capital-letter romaji, too? I don't see very many USTs with capital letters, but if they're common, I should probably take care of it. How common is it?

Hmm, I didn't think of that. Yeah, caps is going to give problems for romaji to hiragana conversion.
I doubt it's common at all. I only saw them when I did some mistake doing input of lyrics... But I haven't seen many romaji ust.

How about to make a different java program for capital romaji to lowcase romaji conversion?
I think to change the main program with this could damage the high compability (the byte to byte comparation). If you do it manually it could work but that seems like too much extra work instead of just using a new program with .toLowerCase().

It could be something simple as (not really coded as I know no JAVA)-:

String temp_string
String changed_string
For(whole text) until end_of_tex{
temp_string = actuallineoftext()
if (temp_string.startsWith("Lyric="){
changed_string="Lyric="+temp_string.substring(6).toLowerCase()
replace actuallineoftext() with changed_string
}
next line}


Or something like that. I'm more of a C++ person and strings are always really picky so I guess it's wrong in the reference vs value issue.
Also R and r being both usable silences helps.

It would damage the byte-byte compatibility, yeah. I do know commands that allow one to ignore case, but I kinda feel weird trying to implement that in a program like this.

I do know that sango312's "Houki Boshi" ust starts with "Yo" exactly cased like that, but I know that's easily fixable since it's only that note that's like that. I haven't seen usts that are capital the entire way through. Maybe I'll wait to see if someone asks about it.

This is so helpful. Thank you so muchhhhh <33333

I sense that I'll be using this program a lot. Now I won't need to download aliased japanese voicebanks.

Oh, and sorry for being so annoying, this program needs java installed in your computer to run, isn't it?

Maybe people without java can't run the program at all (but what person hasn't java today?).

For those people who seem to not be able to run the converter you could insert an advice to go to:
http://www.java.com/en/download/installed.jsp

That's the official English site to check if your java is the latest one and tells you how to update if you're not.

Nowadays Java comes with computers...

The program changes files based on whatever is in a specified text file containing a list of what to change. If you want something different, replace/edit the hiratoroma.txt or romatohira.txt files with whatever you want. The program doesnt have to be changed to do that. Such changes can be made by users editing these pre-existing files in MS Word, for example.

Heck, I'll do it. This is based of the Convert 2.2 files. I put a zip file here with files to convert various capitalizations. 'Master Cap Romaji to Hiragan.bat' will convert romaji to hiragana regardless of whether there are no capitalizations, the first romaji is capitalized or all romaji characters are capitalized. I didnt make it look for odd combinations though. kYO KyO and KYo wont be changed to hiragana, for example. Maybe another bored/helpful persion will add those.

https://sites.google.com/site/bluexgreen/resources
This doesnt include the origional program files. You still need to get those from Aster Selene's link.


EDIT: Another thing you can do is use this to get rid of mixed capitals in a romaji ust. Use the 'master' bat to convert it to hiragana, then use another bat to convert it to all-lower-case romaji.

v2.3 released. Now contains capital-letter support (which I started before BlueboyX, but thank you anyway) along with full OREMO reclist support.

A surprise is coming in the next version!

All right, when I said there was a surprise, I meant it. Read the original post to see what's happened!

You have just won the 120productions (my YouTube/semiorganization) seal of approval. Our seal of approval isn't well known enoff to be flashed around. Lol but you still won one. And your the first official one congratz

Strong work.

There is one more function I can think of. In the oto.ini files you can replace the ,,,,, with something like 10,35,20,20,10 The purpose of doing this would be to give initial values that are easy to modify. When UTAU generates an oto file it just makes the text entry of ,,,,, and the graphical editor basically displays this as having an offset, consonant, preutterane, overlap and cutoff of zero. Cutoff is at the end of thie file, but everything else is right on top of eah other at the beginning. This means that you have to manually seperate everything for every sample if you like to edit them on the graphic interface. This is painful for 170+ sample voicebanks and I have found using the initial values above to make the process a lot easier.

I considered this, but when your settings are nowhere near something like that (as in the case of Hana) or when you don't even need settings at all (as in the case of many voicebanks recorded by OREMO), it could actually be more painful to drag prerecorded settings. Especially when UTAU usually is able to figure out top/end of data by itself.

Oh, and version 3.2 updated. This version is most important, since it fixes a bug in the romaji-to-hiragana aliaser.

ohhhh my gosh it does aliases ;w;

you are amazing

Hi, I've tried to use this forwards and backwards, left to right, upside down and loopdeloop'd, and I could not get this to work. I reinstalled Java twice.

I use windows 7, I've read the instructions over and over, and cannot, for the life of me, get this to work.

Can anyone help me?

Which version are you running?

BlueboyX wrote:
There is one more function I can think of. In the oto.ini files you can replace the ,,,,, with something like 10,35,20,20,10 The purpose of doing this would be to give initial values that are easy to modify.

You say this could be uneeded and annoying BUT
VCVs. It's hell changing the settings for... VCVs. I actually REALLY need it right now, since the oto's I used for the bank don't fit the recordings.

I have to change it (everytime I do an oto) from *insertnumber*,500,-777.777,333.333,111.111 to *insertnumber*,300,-450,222,111.

AKA, I really need this function to make VCVs sound better... Without giving up because it's stupid.

...If you implemented this, I could whizz through VCVs that needed custom settings, Aster. ; A; I'd love you forever.

...pleasethinkaboutitforVCVs?

Since all the samples in a VCV bank have the same Pre-utt, overlap and end cut off settings, it would be really useful XD You wouldn't have to change any of that around.

I would like to do that, but the problem is that since I use a general search-and-replace algorithm, one would have to blank all of the settings to do that (because once the settings are in, you cannot change them).

Aster Selene wrote:I would like to do that, but the problem is that since I use a general search-and-replace algorithm, one would have to blank all of the settings to do that (because once the settings are in, you cannot change them).

Hm, Search and replace...
So there's no way you could have the user be able to enter the settings you need to replace?
(Like, if I open my oto settings in notepad, set it to change all the "500,-777/777,333.333,111.111"s in the notepad file?

I'd be willing to blank it out if necessary, though.

Sadly, I can't do that - the only thing I can do is put in preset settings (which will only make your day worse).

*cries*
There's a way to do it, just not like this...
Goddamn. XD I might have to dabble in scripts and coding today to try and find a way to batch change these files.