Changes on ES VCCV phonemes

[jakisalt]

Acknowledgement

• I have read Getting-Started and FAQ
• If dealing with version 0.1.549 or higher, you have installed the .NET 8 runtime

🐛 Describe the bug

This phonemizer improved a lot on this update and I really liked it. However, it now converts certain words and syllables into phonetic aliases that are completely different from the original nJokis phonetic system. For example:

be -> [-Be] (before: [-be])
lle -> [sxe] (before: [jje])
hoy -> [- o] [oy] (before: [- o] [oI])
causa -> [-ka] [aw] [ws] [sa] [a-] (before: [-ka] [aU] [Us] [sa] [a-])

I tried switching between different nJokis-based Spanish voicebanks on the same .ustx (with ES VCCV already set as the phonemizer) and each one of them was having the same issue; they couldn't play certain sounds/notes since the new aliases from this update weren't found in their oto.ini files.

Another error shows up when switching to another voicebank while using this phonemizer. This doesn't seem like an oto.ini issue since 1: this happens with all my Spanish voicebanks, 2: all of them worked properly with previous versions of ES VCCV. Besides, this error continues showing up even if I open a different sequence file!!

A downside I see from this is I genuinely doubt future Spanish voicebanks based on this recording method will adapt to this new dictionary taking into account the official nJokis phonetic chart still has the same phonetic system that previous versions of ES VCCV had and people will continue using it, breaking the compatibility with the new aliases/dictionary a bit.

1: Is there any workaround for this (besides manually editing each phoneme) so this phonemizer can convert the input text using the dictionary from previous versions again?
2: Where did these new aliases come from?

Thx for your attention c:

Explains how to reproduce the bug

"Oto not found for [insert phoneme]":

1. Open .ustx file and choose a nJokis Spanish voicebank
2. Set phonemizer to "ES VCCV"

"Index was outside the bounds of the array" (all phonemes show "error"):

1. Open .ustx file and choose a nJokis Spanish voicebank
2. Set phonemizer to "ES VCCV"
3. Choose a different Spanish voicebank
4. Switch back to the previous voicebank

OS & Version

EndeavourOS (Linux 7.1.4-arch1-1)

Logs

log20260908.txt

[HariZaKitsukami]

Adding some others I've noticed since I also actively use Spanish:

• Any plosive in VC -> t even when the corresponding alias exists
• ch -> cx (it and jj -> sx likely a result of jj -> sh and then h -> x)
• E (semivowel) -> e (full vowel)
• d and g have the same issue as b

[HariZaKitsukami]

Oh, regarding question 2.
All these issues seem to be the result of replacements made in ValidateAlias() (which basically makes a bunch of replacements to the alias when it isn't found in hope of getting a new alias that can be found). Either the replacements are being applied indiscriminately, or something is making it impossible to find the first alias, therefore necessitating the validated alias.
Sometimes it's necessary (e.g. to avoid intervocalic [b] [d] [g], for voicebanks using [sh] for [jj] and/or [h] for [x], and to account for seseo), but in this case it's making the changes even when unnecessary.