Disappointed of Korean XLS, honestly

Synthesizer V Studio
1

As a South Korean myself, I was very hyped for the Korean XLS. However, it showed some flaws.

The biggest flaw is from the s and s_t phonemes, which stand for Korean ㅅ and ㅆ, respectively. By the Korean phonetics, these phonemes are supposed to assimilate into the voiceless alveolo-palatal fricative, or [ɕ] in IPA, when followed by i or j. Though Uni expressed this assimilation well in her demos, other (non-Korean) voicebanks don’t do that well.

Here’s my proposed remedy:

  • Add s\ and s\_t phonemes for palatalized ㅅ and ㅆ, respectively.
  • Add the H (or [ɥ] in IPA) phoneme. With this phoneme, ㅛ, ㅠ, and ㅟ should be H o, H u, and H i respectively.

What are your opinions on this?

2 Likes
2

I don’t have a Korean library, but I was rather surprised at how poorly the libraries I have performed in Korean. With other languages there’s a clear accent when they sing in their non-native language, and that’s to be expected, but the phonemes at least are correct. I don’t feel like that’s the case with Korean.

1 Like
3

In Spanish some phonemes are sounding wrong too. Hope they will fix the different languages soon.

1 Like
4

The accents in XLS sound worse in SynthV 2. Some reddit user proposed having an “accent slider”, and I think this is a good idea.