Overview

So you’ve bought SynthV and you just happen to be a conlanger, right? You want to force your e-waifus to sing in your conlangs (or someone else’s), right? Consider this the guide for you, then, you sicko.

(Disclaimer: This guide might sound pretty blunt, but remember that this banter also applies to me.)

Which voicebank language to use?

English and Japanese are the main languages we’ll be focusing on. Mandarin Chinese is also another option if you have some money to chuck at a paid voicebank.

English

Voicebanks: Eleanor Forte.

Strengths:

• Several consonants not found in Japanese voicebanks: /θ ð ɹ l/
• Good consonant cluster handling

Weaknesses:

• Has a lot of trouble producing anything close to /a o/
• Same vowel symbol can produce different results in different places
• /l/ is actually [ɫ]
• /r/ can be emulated but hard to find the right vowel
• Aspirates voiceless plosives, which might or might not be what you want

Japanese

Voicebanks: Yamine Renri, Genbu.

Strengths:

• Has (almost) the classic five-vowel system, with consistent results
• /r/ can be emulated easily

Weaknesses:

• Doesn’t have any vowels beyond the classic five-vowel system
• Doesn’t have /θ ð ɹ/ (or /x ɣ ɬ ɮ q ʕ/ &c. for that matter, but neither do any other voicebanks)
• Often has trouble with complex consonant clusters – one final consonant is (surprisingly) fine, but has trouble with things such as /ns/ in the coda

Mandarin

Voicebanks: AiKO, Chiyu, Shian, Cangqiong, Haiyi

Note that this is pretty much wild speculation since there aren’t any free voicebanks for this language as of 23X2019.

Strengths:

• Lots of vowels: [a ɑ o ə e ɤ ʊ u i ɨ i˞ y] for the monopthongs
• Has both [ɻ] and [l]
• Has both aspirated and unaspirated stops

Weaknesses:

• No voiced stops
• Some sounds such as /v/ are missing too

But I’ll focus on English and Japanese VBs here.

Two or more of the above

Combining voicebanks of multiple languages can ‘iron out’ each other’s weaknesses. See this, this or this video for examples.

If you have two tracks playing at the same time in general, you can also use different phonemes for each track.

Consonants

The easy ones

/p b m f v t d n s z j k ɡ w h/

• E: p b m f v t d n s z y k g w h
• J: p b m f v t d n s z y k g w h

Technically, f in Japanese represents [ɸ], but what kind of sicko language distinguishes /ɸ f/? (I’m looking at you, Lineparine.)

Genbu has issues pronouncing w e, w i and w u at the start of a phrase, pronouncing them with something like [ɹ] instead. This can be exploited for conlang purposes, but if you really want /w/, prefix a 1/16-beat note with w o then take the onset off the main note.

/ʃ ʒ tʃ dʒ/ and friends

English has sh zh ch jh for these, which represent [ʃ ʒ tʃ dʒ].

Japanese has sh ch j for all but the second, which represent the respective alveopalatal consonants. /ʒ~ʑ/ is the hardest to emulate. (Though Genbu’s j sounds more like /ʒ~ʑ/.)

/ɾ r/

For the tap, use dx for English or r for Japanese.

To get the trill, put a few short notes before the main note with dx ax or r u, then start the main note with r.

/l/

English has l for this, easy peasy if you don’t mind having it dark. Sadly, SynthV lacks any equivalent to Vocaloid’s l0.

Japanese has no l, so you’ll have to use r instead. If you have one in a coda, increasing the Syllable Coda property of the note can make it sound more like an /l/.

/ɹ/

If you’re a good and impartial conlanger like I am not, then you shouldn’t need to read this section very much – in fact, you should expect to read this once for every fifty conlangs you make.

On the other hand, if you just loooooooove the sound of /ɹ/, then English has r, right at it if you don’t mind some ‘contamination’.

If you’re using Genbu, then w works before e, i or u at the start of a phrase. Most of the time, though, you’re going to have to settle with r. r w might be a better option sometimes.

If you’re forcing Renri to sing with Genbu, then try having one sing r and the other w.

/θ ð/

Again, you shouldn’t need this section much if you’re impartial, though including one of these phonemes in your conlang’s inventory isn’t as big of a sin as including /ɹ/.

But if you just looooove how <þ> looks, then English has th dh.

Japanese gets the short end of the stick – the closest you have is s z, which is a terrible approximation.

If you’re layering multiple tracks, avoid having s or z if you have English th or dh elsewhere, or it will be drowned out. Consider using d for /ð/ or omitting the consonant altogether.

/ŋ/

Now this is one sound everyone should put in their inventories and allow in the onset.

English uses ng for this.

Japanese has N, but it’s treated as a vowel, so when used as-is, it’ll take too long to pronounce. Move it to another note so you can control the length.

/ɬ tɬ/

Hey, at least /ɬ/ is more common cross-linguistically than /θ/ or /ð/!

This sound isn’t in English or Japanese VBs (Welsh VB when???), but luckily, it’s easy to emulate.

For English VBs, use s l in the onset position or l s in the coda position. Then set Syllable Onset to -1.5pow2 for the former, or Syllable Coda to +1.5pow2 for the latter. For /tɬ/, add a t to the front.

For Japanese VBs, use s r in the onset position. (Unfortunately, this only works for onset positions in this case.) Again, set Syllable Onset to -1.5pow2. For /tɬ/, use ts in place of s.

It should be noted that this trick works best with Eleanor and Renri. Genbu doesn’t get very good results from it.

I don’t know any way to get /ɮ/.

/x ɣ/

If you’re layering two voices, have the top one pronounce hh or h and the bottom one pronounce k or g. You might need to adjust the note timing to get the consonants to be in sync.

Alternatively, cocoa found a way to pull off an /x/ (example s5p):

it’s just [hh hh] with a mild scream, and crazy volume breathiness and voice

Vowels

If you’re covering a Drsk song, then you don’t need to read this section, but otherwise…

/a/

For Japanese VBs, a does the trick.

Unfortunately, it’s not so easy for English VBs. The first choice is ah (/ʌ/), but that doesn’t always work and you might need to try out aa, ae or ao, or even ay or aw if you’re unlucky. It should be possible in theory to use aa and adjust the Gender parameter slightly towards the feminine end to raise the F1 and F2 formants (thanks, paradoxelle).

/i/

i for Japanese VBs, iy for English VBs, easy.

/u/

u for Japanese VBs (though it’s less rounded than the standard-issue /u/), uw for English VBs.

/e/

e for Japanese VBs, full stop.

For English VBs, eh works most of the time, but if that’s pronounced too closedly, try ae as well.

/o/

o for Japanese VBs.

For English VBs, ow is the standard choice. Up the Syllable Nucleus property to turn down the glide. Use ao before l or r, but it doesn’t really work well elsewhere, at least not with Eleanor.

/y/

Boy, have you landed yourself straight in hell!

y u is the closest I can get with Japanese.

For English, the closest is w iy.

/ə/

Azur Lane reject can just use ax. Star Girl and Rope Boy, on the other hand, are out of luck.