Resource Vocal Synth Voice Characteristics

[Wario94]

Just curious, could we talk about the web version of SynthesizerV on this forum?

 

[Scarlet Illusion]

Yes, of course! Information about web SynthV is more than welcome! :D

 

[Scarlet Illusion]

I think I'll mention something I noticed, but I'm not going to add it to the OP just yet. I'm working on a cover with Teto that's out of her range (it's too high) and causes a lot of the notes to sound shrieky, and so I tried a two other UTAU: Ruko (male, kire/power) and Tei. Although the notes are well within Ruko's range, he sounds quite shrieky, even after I change the prefix. The voice also begins cracking a little when out of range, and the voice begins collapsing, making it hard to hear the voice actually pronounce anything. However, I switched the vocals to Tei, and, although it still sounds a little shrieky, you can actually hear the voice pronounce the syllables this time. Keep in mind, Teto and Tei's recommended ranges go up to B5 while Ruko's goes up to about D#6. Maybe this has something to do with the style of Tei's voice compared to Teto and Ruko's voices (Tei sound higher pitched and is much softer than Teto and Ruko)? Also, she averages D4 for her freq. average while Ruko averages F4 and Teto average around E4.

Maybe one of you all have an explanation for this? It's not something I need an answer on, but I am curious why this would be.

 

[___]

Maybe this has something to do with the style of Tei's voice compared to Teto and Ruko's voices (Tei sound higher pitched and is much softer than Teto and Ruko)?

That's most likely it, from my experience it takes less for a shouty voice to collapse as opposed to soft/normal voice. Some of the utaus I use often are very shouty in their high range and I have to deal with this often.

 

[Scarlet Illusion]

That's really interesting! I would have thought that a softer voice would collapse quicker due to not being very strong.
----------
Also, unrelated note here: I noticed that Mayu seems to be another case of added nasaliness by the Vocaloid engine. From what I've heard of her voice provider, her vp isn't very nasally. (Also, anyone's welcome to fight me on this, but Mayu doesn't sound that much like her VP even though she's supposed to be the Vocaloid version of her VP...)

 

[peaches2217]

I've had her for thirty whole minutes and I've already got some tea on Haiyi.

• I'd pin her range at around G#3-A#5, with the middle of that range being ideal.
• Around D5, her voice loses volume, but not power.
• She responds poorly to Tension adjustments in her higher register, but otherwise responds well. (Making very high notes sound right is probably an issue that'll need addressed in post.)
• When singing a long note (usually three measures or more), her volume will dip at the beginning of the note and crescendo, a lot like singers of traditional East Asian music. (This doesn't seem to apply when holding out a single vowel, like a, but it does when singing an actual word.)

 

[Scarlet Illusion]

Thanks for the tea, sis! XD

I've added it to the OP!

 

[peaches2217]

I found a fix for her high notes!

• Setting the note to C4 (right-click the note, click "Edit Sub-Library," and chose C4 from the drop-down) will help keep higher notes consistent with the rest of her notes.

 

[Scarlet Illusion]

Duly noted! Thank you!

 

[Scarlet Illusion]

Alrighty. *cracks knuckles*

I messed around with Ameya's new resampler last night and here's some things I found:
Doppeltler softens Tei's voice a lot and makes it weaker, while Moresampler makes her voice stronger and deeper and might be giving her a thicker timbre?

You could exploit this difference to give the voice more versatility. If you want to have an UTAU sing a softer song, you could use Doppeltler, but if you need more strength in voice, you could use Moresampler. Idk how if these things effect other UTAUloids. So far I've only tried with Tei.

 

[peaches2217]

Here's some info on Iroha (specifically V4 Natural)! It's not much, but it's what I've got.

• Her vowels tend to sound nasally, especially on held notes (I'd pin it as slightly less nasally than Gumi Native).
• Her plosives are very loud and a bit harsh, especially [t]. Even turning the accent down to 0 doesn't help much.
• Her [h] and her fricatives, likewise, sound kind of harsh.
• Has a very lovely alto range. Her voice retains just as much power in her lows as in her highs.
• Risks sounding flat in her lower range. This can be fixed by raising PIT a few points.

 

[Scarlet Illusion]

Thank you for your contribution! I've added it to the OP.
-----
Additionally, I've updated Tei's section a little.

 

[Cerāgi <º•]

*UPDATED*
Soo, this days I’m messing around around with Luka V4x’s trial, and even though she’s a bit difficult to use, I’m really loving her and I think she has a lot of potential , so I wanted to write here something I noticed about her quirks, hoping for them to be somewhat useful (I’ve only used her trial for a few days though, so if anyone more experienced than me using her(so literally anyone lol) would like to correct me in anything feel free to do it ahah



So let’s start, I’ll divide this in two sections in order to make it easier



A) Luka V4X Hard

B) Regarding EVEC



A) Quirks and workarounds

• She has choppy transitions, using [-] notes helps in connecting notes better as well as correcting the timing of the transition itself; using Pitch Rendering/Piapro automatic one can be really useful, as it lets you directly see when the pitch is actually changing and “going” to the next note
• Her high notes can sometimes sound distorted, applying an EVEC color usually solves this
• Her consonants can sound too “hard” at times (most notoriously the “k” in [k a] and “t” related syllables (ex. [t a]), changing to the voiceless one or to the matching voiced consonant makes it less prominent (ex. [k_0 a], [g a])
• For V-V transitions especially, but for other notes too, using a small “soft” or “dark” EVEC helps in smoothening them out (ex. [a][a#2/a#-]); this can help even when using different EVEC colors together (ex. Cute EVEC(a)[a a#2/a#-][Whisper EVEC(i)]
  [*]When using notes that are too short (usually with 1/64 ones) next to each each other she can sometimes pronounce the consonant of a following note in a very loose, almost voiceless way; should this effect be undesired make the previous note a bit longer, leave a small pause between a note and the other or change the consonant to another one that is close in sound (ex. [a][k a] ->[a][g a])
  [*]Be careful when increasing DYN, as it can easily make some notes greatly louder than others, making her glitches a lot more apparent, on top of that
  [*]She benefits from some BRI added to the vocal, but using too much can lead to distortions and glitch sounds
  [*]Adding CLE to “singer settings” seems like it can easily add a strange metallic sound even when used in small amounts
  

B) EVEC notes

• • • Sometimes they can sound not as effective as they should, in this case just dividing the EVEC notes and putting a small“dash” in between make them stand out more (ex. (Cute EVEC)[a] —> [a][-][a#+])
    • Sometimes, especially on longer notes with vibrato applied, but also on some high notes, mostly POWER 2, CUTE and sometimes POWER 1 EVEC distort and generate an unwanted warp noise; making as follows seems to usually solve this: (ex (Cute EVEC)[a] -> [a][- a#+](Cute EVEC)[a]). If even this doesn’t work try to play around with [-] and try some different combinations of EVEC colors
    • If POWER 2 sounds distorted, swapping to POWER 1 can sometimes solve this
    • If choosing FALSETTO EVEC, previously using a small WHISPER EVEC note helps in changing to this register in a less drastic way, making it more controlled and smoother overall
    • Using some GWL at the very beginning of a note can give a little more power to it that sometimes the two POWER EVEC cannot show properly
    • If EVEC to EVEC note transitions sound choppy, just follow what have been said to the previous section (A) and you should be fine (usually though, small [-] notes are helpful in this)

 

[Scarlet Illusion]

Soo, this days I’m messing around around with Luka V4x’s trial, and even though she’s a bit difficult to use, I’m really loving her and I think she has a lot of potential , so I wanted to write here something I noticed about her quirks, hoping for them to be somewhat useful (I’ve only used her trial for a few days though, so if anyone more experienced than me using her(so literally anyone lol) would like to correct me in anything feel free to do it ahah

Thank you for the information!! I quoted it in the original post!! :D

 

[Prism]

So I noticed something with Ruby that when you render her in v5 she's a lot less nasally and glitchy than v4. I've noticed that the v5 rendering engine handles nasal and glitchy voice types a lot better so if you have a problem with v4 try v5

 

[peaches2217]

I'm not sure if this is exclusive to Luka V4X or no, but she has a tendency to sound flat. Raising her pitch a tiny bit fixes it pretty well. (I usually set her between +599-799.)

 

[Scarlet Illusion]

I'm not sure if this is exclusive to Luka V4X or no, but she has a tendency to sound flat. Raising her pitch a tiny bit fixes it pretty well. (I usually set her between +599-799.)

So I noticed something with Ruby that when you render her in v5 she's a lot less nasally and glitchy than v4. I've noticed that the v5 rendering engine handles nasal and glitchy voice types a lot better so if you have a problem with v4 try v5

Thank you both for the info! Prism, I've quoted your info in the OP! Peaches, your information has been added to the OP! (Also, I'll have to try using that idea with Miku. Miku almost always sound flat for me. :P )

 

[peaches2217]

Can I make a note about SynthV in general, Genbu in particular? When a vocal (especially Genbu) pronounces a hard consonant (k and t especially), the note that comes before it will often be cut short. This can be fixed by lowering the Syllable Onset of the note with the consonant.

I'm not sure if that made sense or not. Forgive me.

 

[Scarlet Illusion]

Can I make a note about SynthV in general, Genbu in particular? When a vocal (especially Genbu) pronounces a hard consonant (k and t especially), the note that comes before it will often be cut short. This can be fixed by lowering the Syllable Onset of the note with the consonant.

I'm not sure if that made sense or not. Forgive me.

That makes perfect sense! I've added it to the OP! Thank you.

 

[peaches2217]

I figured out the words I was looking for!! The Syllable Onset functions in the same way Velocity does in Vocaloid. That might be a less wordy way of putting it. Thank you!