Anyone think that this reclist is worth it?

socialgutbrain777

Teto's Territory
Defender of Defoko
I've been working on a reclist for months on end and my biggest struggle was figuring out phonemes.​

Now, I'm aiming for a Multilingual type of deal because efficiency and not having to record separate voicebanks for your UTAUloid for every language you want them to sing in is nice. I considered using a pre-existing dictionary, but it wouldn't make sense as none of the phonemes would work with romaji. So, I came up with my own set of phonemes for my vowels, as seen below:

a as in box
i as in free
u as in news
e as in end
o as in kyo
@ as in trash
6 as in book
3 as in birds
7 as in lip
5 as in love​

In hindsight, it bothered me because the number-vowels felt kind of off-putting, mostly because someone might have a suffix for their voicebank that's a number, resulting in some confusion here and there.

My other phoneme set had Arpasing vowels for the English-y vowels, which were capitalized to avoid confusion. They felt more comprehensible, but still, I couldn't decide if I wanted my diphthong vowels to be typed as "AY" or "ai", and the pseudo-paradox started to get to me to the point where I noped out of that one.

Finally, after so much trial and error, it came to whatever this thing is:

a as in box [ba a_k k_s]
i as in free [f_rh rhi i_-]
u as in news [nu u_z]
e as in end [-e e_n n_d]
o as in kyo [kyo o_-]
& as in trash [t_rh rh& &_sh]
@ as in book [b@ @_k]
X as in birds [bX X_d d_z]
! as in lip [l! !_p]
C as in love [lC C_v]
r as in roki [ro o_k ki i_-]
rh as in roll (English R) [rho o_@ @_l]

An Example of a sentence in these phonemes: I'M A GIRAFFE!!!

[-a a_! !_m mC C_j jX X_rh rh& &_f]

Another example: おはようございます。 (Ohayo gozaimasu!)

[-o o_w ha a_i i_o o_g go o_z za a_i i_m ma a_s u_-]

And yet another example, because I just now realized that doing this is surprisingly fun: Look at the sky tonight, darling!

[l@ @_k k& &_t t_dh dh@ @_s s_k ka a_i !_t tC C_n na a_! !_t R da a_rh rh_l l! !_ng]
None of the English vowels are upper-case vowels, as file names aren't case sensitive (I'm aware that something like a dash is added to the name to differentiate them), and my brain gets confused sometimes when viewing them. Now, of course, this looks confusing, because the letter c is typically used to represent the vowel in cot, not the vowel in love. But, since this reclist is meant to be rather light, and therefore merges the box and cot vowels, this frees up a bit of room to use this capital C as something else.

Anyway, with that being said, does anyone think that this is really worth it? CV-C, which is a fantastic Multilingual reclist released earlier this year, I personally like, and the English sounds ridiculously smooth. I would've made my own version with my own phonemes, but I'd feel bad for butchering an already glorious reclist. I almost feel like dropping my reclist faster than you can say omae wa mou shindeiru. I'd still like some input, though, so I can hopefully do something right.
 

Kiyoteru

UtaForum power user
Supporter
Defender of Defoko
If you're planning to do a multilingual reclist, I would highly suggest using an encoding like X-SAMPA, which directly corresponds with the International Phonetic Alphabet. Even if you only focus on one or two languages now, an established system will give you the right amount of room to expand in the future.

Even so, one of the problems with multilingual banks is that a phoneme written the same way in two different languages may actually be pronounced slightly differently. For example, this is the "canonical" vowel chart
upload_2019-9-2_12-41-57.png
And this is the vowel chart for Swedish.
Swedish_monophthongs_chart.png


Attempting to reuse vowels recorded perfectly for Swedish in another language will result in, of course, a Swedish accent. This is one of the reasons that languages will be recorded as completely separate sets of samples. But if this is a sacrifice you're willing to make, best of luck.
 

Similar threads