|!||The following is a tutorial made for VOCALOID fans by fellow VOCALOID fans.||!|
Japanese VOCALOIDs are VOCALOIDs that are capable of mimicking the Japanese language much easier than VOCALOIDs of other languages. The followings are lists of phonemes needed to make the VOCALOID sing in Japanese.
The origin of the language is mostly unknown, including when it first appeared in Japan. Its main influences are Chinese and Old Japanese. More modern decades have seen many European influences on the language, especially many English loanwords having been adopted into the Japanese phonetic system. However, the lack of influence from other languages, in addition Japan's isolation from the rest of the world, has contributed much to the precision of the Japanese phonetic system.
Far less new sounds entered the language for many centuries, in comparison to other ones such as English which had heavy influences from other languages.
VOCALOID and the Japanese LanguageEdit
Due to lack of influence and the isolation of the language, this has worked in favour of the language within VOCALOID. The result is that Japanese voicebanks are some of the simplest to make for VOCALOID, as the sounds simply have to be correct and errors in Japanese language skills often are minimum, being a result of general bugs, glitchs and the odd mispronunciations. The language is fairly straight forward to produce as most sounds are more definite than with languages such as English.
In addition, sounds such as diaphones and triphones often get acquired from recording data used for the languages main sounds, as a result when triphones were introduced in VOCALOID3, old voicebanks already had recorded the necessary triphones. This made updating the voicebanks fairly easy from VOCALOID2 to VOCALOID3 or VOCALOID4. In addition, development occurs significantly faster due to the lesser amount of sounds required. Japanese is one of the cheapest and fastest languages to produce for VOCALOID overall, with even multi-voicebank releases seeing their provider spend no more time in the studio than approximately a week.
One downside to the language is that during recording, less overall traits can be captured with certain recording techniques. This leads to Japanese VOCALOIDs at times having very little variation in how they sound. This especially impacted VOCALOID3, as this engine introduced the largest number of new Japanese voicebanks and common "vocal types" began to form especially among the female voicebanks. A new recording technique was introduced for VOCALOID4 by Internet Co., Ltd., allowing to capture more traits. As a result, VOCALOID4 and later Japanese voicebanks often do not behave the same even when they sound similar and have far more quirks and characteristics then pre-VOCALOID4 ones.
Another issue is that when a Japanese voicebank says a certain word, there is no work around like with larger language voicebanks. In addition as a contrast to languages as English or Spanish Phonetics, there is less overall variation of sounds, which means the manner a voicebank says a certain sound will often be identical every single time. There is simply no need for many variable of certain sounds such as "ma", "do", etc, so far less diaphones and triphones are needed overall, with some sounds being largely the same if they are at the beginning, middle or end of a word.
While this means the Japanese voicebanks are often among the most consistant and stable, they can suffer due to the repetition of their limited number of sounds. They can often be easier to edit, fix and adjust because of this overall because the result of the limitation variation is much more consistency's and predictabilities within each voicebanks. Once a producer learns something, such as the weakness of the sound "ma" in Hatsune Miku's VOCALOID2 vocal, you can easily pick up on such as weakness every single time. In other languages, though it may be common for "ma" to be a weakness, it may not be a weakness in every example, making them much less predictable and consistant.
This makes them also fairly simple to tune as a result. Often it is possible to use VSQ, VSQX or VPR files for any Japanese Vocaloid, with sometimes only slight tweaks here and there overall. However, that does not mean a user does not have to check all vocal results at all for any weakness in the voicebanks vocal performance and pronunciation, as that is never advisable at all.
Japanese VOCALOIDs can use a standard YAMAHA script.
The script has been adjusted considerably since the early VOCALOID days, with adjustments even being made during the production of the Hatsune Miku vocal in VOCALOID2 to improve the needed sounds for Japanese. Thus, there was a slight leap between VOCALOID and VOCALOID2 quality even without the engine itself taken into account. With the addition of triphones in VOCALOID3, Japanese Vocaloids also became much smoother then VOCALOID2 vocals.
Notes on AccentsEdit
Despite the general belief that singers completely lose their accents when they sing, this is not the case in every instance and an accent is possible to be heard even in singing vocals.
However, the reason many are led to believe this is that there are several methods of training singers to disguise or otherwise hide their natural accents - they may even adopt an accent that isn't their own for singing. Samples include genres such as western or country, black music such as Jazz or Soul. Singing also uses different muscles to speech, resulting in difference of air pressure and way the throat moves. Genres such as Opera are most likely to make a accent appear almost entirely absent thanks to the impact of the opera vibrato.
VOCALOID will capture any form of accent quite easily at times. It depends on the recording method used on the voicer, type of sound being recorded per sample (accent impact varies per sample and language) and overall number of samples that make up the voicebank (the more samples, the more chance of it slipping in).
For Japanese Vocals, accents can appear in a voicebank, but their impact is very little on how it sounds and has a tendency to impact non-native Japanese voicebanks rather then the native Japanese voicebanks.
The reason being is that Japanese mostly uses pitch accent and words do not need to be stressed at all. Thus, in the majority of cases, the typical Japanese VSQ/VSQX file will work without little adjustments. Any oddities in regards to how a vocal sounds tends to be a result of a fatal flaw in the voicebank and its samples themselves or with VOCALOID, rather then a result of a form of accent. However, one important element to note that an accent can be a contributor to the common issue with Japanese voicebanks. If a sound is incorrect due to an accent, then the language itself can often be off. As with all other cases, there is rarely a alternative sound that can be used to replace such cases.
In general, accents tend to serve only as a contributor to the "traits" of a Japanese voicebank, giving it a little bit more distinction at times then others and can be dismissed as such except in extreme cases. As previous mention, due to the limited number of phonetics it is much harder to record the traits of a vocalist provider, accent as well may be impacted just for this reason alone in addition. As such, all os these things can cause the accent to have very little impact on the performance of the voicebanks overall compared to the accents captured on voicebanks of other languages. Quite often, these simply add to the voicebanks performance traits and not the language traits.
- Hatsune Miku
- Kagamine Rin
- Kagamine Len
- Camui Gackpo
- Megurine Luka
- SF-A2 miki
- Kaai Yuki
- Hiyama Kiyoteru
- Nekomura Iroha
- Utatane Piko
- Tone Rion
- Yuzuki Yukari
- Aoki Lapis
- Macne Nana
- Tohoku Zunko
- Otomachi Una
- Macne Petit
- Yumemi Nemu
- Masaoka Azuki
- Kobayashi Matcha
- Kizuna Akari
- Mirai Komachi
- Haruno Sora
- Meika Hime
- Meika Mikoto
The following are VOCALOIDs who were confirmed to be sampled from multiple voice providers for their Japanese voicebanks.
Phonetic System's CharacteristicsEdit
There are 41 phonetic pronunciations which make up the Japanese Vocaloid library, these phonetic inputs will use any set of the estimated 500 total samples needed for Japanese recreation per pitch.
Due its moraic nature, the Japanese language has a simple phonotactics and syllable structure. For this reason the Japanese Phonetic system was designed to be encoded as [C V] syllables. For that reason, the voicebanks may struggle in pronouncing consonant clusters, diphthongs or consonants in coda position.
The Japanese Phonetic System includes the 5 vowels of the Japanese Language.
As per the palatalization phenomena found in the Japanese Language, the system is designed so that the vowel [i] needs to have a palatalized consonant in front of it to produce sound. If this isn't the case then the combination will be silent, even if both phonemes are separated in different notes. The only exception to this are the phonemes [s] and [dz], as those ones produce sound when followed by an [i].
It's important to note that some voicebanks may have problems with certain vowel combinations, which can end up sounding choppy. Some techniques to help correct this exist. Generally, this was a more common problem in the first generation of the software but as the release of the Japanese voicebanks progressed, the vowel combination problem become much less apparent. This is due to the improvement of the recording and processing methods aswell as overall experience with the synthesis engine within the companies. The problem was phased out completely by the third generation.
The Japanese Phonetic System includes 36 consonant phonetic pronunciations. Due to Japanese being a language which has little to no consonant clusters, the system was designed without consideration to standalone consonants. Because of this, consonants always need to be accompanied by a vowel. If not, the synthesizer won't be capable of reproducing the consonant. This will instead generate audio distortion, clicks, electronic buzzing or sound loops.
The exception to this are the nasal consonants associated to the Japanese N or ん, which is the only consonant in the Japanese Language which is pronounced without a vowel, as this character is considered a mora in their own.
In the case of the consonants, due the Yōon and the related palatalization phenomena of the Japanese Language, the system includes two versions of the same phoneme: the standard one and the respective palatalized version.
The palatalization has two definitions, a phonetic one and a phonological one. For the phonetic term, it refers to a secondary articulation, which adds a small y-like glide sound at the end of the consonant. The phonological term refers to a kind of sound mutation or assimilation process, that changes the sound of a consonant into a more palatal articulation. In the case of the Japanese language, both kinds of phenomena can be found.
For the Japanese Phonetical System, most of the palatalized consonants are differentiated from their standard version with the addition of a small apostrophe ('), which is the X-SAMPA's equivalent to the IPA's small superscript ‹ʲ›, used to denote the secondary palatal articulation. Example: [tʲ] for a palatalized [t].
In the Japanese language, one of the few consonants that are pronounced is the N (ん in hiragana, ン in katakana). This letter has a lot of assimilation allophones, and all those are nasal consonants. Due this, all the nasal phonemes ([n], [J], [m], [m'], [N], [N'], [N\]) can be reproduced standalone, without a vowel accompanying them.
Due to the way the Japanese voicebanks were recorded and the way the Vocaloid editor was made, there are some phoneme combinations that are forbidden or aren't recognized by the synthesizer. If you attempt to enter these combinations they won't produce sound due to the synthesizer not allowing them.
Some of there forbidden combinations are:
- non-palatalized phoneme + [i] (Exceptions: [s], [dz])
- [w M], [j i] and [h M]: nonexistent in the Japanese Language. The [h M] combination is replaced by [p\ M]
- Some palatalized phonemes + vowel different to [i] (check the previous chart)
Also, there are some consonant phonemes that are restricted to certain vowels. If the combination isn't the correct one, the synthesizer won't produce sound.
- [h\]: Restricted to the vowels [e], [o]
- [z] and [Z]: Restricted to the vowels [e], [o], and [M]
A new set of phonemes was added with the release of the VOCALOID3 software. This new set of phonemes are unvoiced versions of the vowels and the sonorant consonants (Liquids and Nasal Consonants, including they palatalized versions) found in the Japanese Phonetic System.
In linguistics, voicelessness is the property of sounds being pronounced without the larynx vibrating. Sometimes the sonorants (vowels and sonorant consonants) can became pronounced in a voiceless manner. When this occurs, you can actually see the person articulate the sonorant, but it's either barely audible or silent altogether.
- Example: the Japanese word sukiyaki is pronounced [su̥.ki.ja.ki]. This may sound like [s.ki.ja.ki] to an English speaker, but the lips can be seen compressing for the [u̥]. Something similar happens in English with words like peculiar [pʰə̥ˈkjuːliɚ] and potato [pʰə̥ˈteɪtoʊ].
To use them, the user must add the suffix [◌_0] to the sonorant, which corresponds to the X-SAMPA's diacritic for <◌̥>, the IPA's diacritic for a voiceless phonation.
- Example: For a voiceless [o] the user must type [o_0]. For a voiceless  the user must type [4_0]
When a VOCALOID2's Japanese voicebank is imported to VOCALOID3 this new set of phonemes is generated from the samples existing on it.
[a]; [e]; [i]; [o]; [M]
[a_0]; [e_0]; [i_0]; [o_0]; [M_0]
[n]; [J]; [m]; [m']; [N]: [N']; [N\]
|[n_0]; [J_0]; [m_0]; [m'_0]; [N_0]: [N'_0]; [N\_0]|
|Liquids||; [4']||[4_0]; [4'_0]|
Fixing choppy vowel combinationsEdit
Is possible to correct the problem of certain vowels combinations that sounds chopped with the aid of the phonemes [j]. [w] and [h\].
The consonant phonemes [j] and [w] can be utilized as semivowels or glides for the vowels [i] and [M] respectively, which allows use them to fix the vowel combinations with those vowels.
These consonants can be utilized either in replacement of their vowel:
- The first Japanese VOCALOIDs (MEIKO and KAITO) have some problems pronouncing [a i]. This can be fixed replacing the [i] for a [j]. [a i] → [a j]
or can used to unite the both vowels inserting it between them (don't forget the combinations [j i] and [w u] are forbidden).
- If the combination [M e] sounds choppy, the note can be split in two . [M e] → [M w][e] or [M][w e] (probably you will need decrease the accent or attack to got a smooth pronunciation)
In the case were you can't use these phonemes, you always can use the restricted phoneme [h/]. This phoneme just produces sound if is succeeded by a [e] or [o], when combined with the other vowels this consonant won't produce any sound. However, if after the mute combination you add a vowel on a different note, the synthesizer will skip the mute combination and immediately will reproduce the following vowel, allowing you fix choppy vowel combinations.
- Miku is known for struggle with the [e] and [o] vowel combinations. When . [o a] → [o h\ a][a] or [o][o h\ a][a]
- The Kagamine Rin / Len ACT2 are known to have various choppy vowel combinations. Due their [h\] is mute with any vowel, this one allows fix any choppy vowel combination.
In VOCALOID2, the phoneme [Asp] generates a similar effect to the phoneme [h\] with any vowel combination, allowing use it with choppy vowel combinations.
Gemination and Consonant LengthEdit
The gemination (consonant length) is when a spoken consonant is pronounced for an audibly longer period of time than a comparative short consonant. This is an important distinctive phonetic process in the Japanese Language.
- Example: Two words can have a different meaning just for the different consonant's length
- 河川 kasen IPA:[ka.sẽɴ] 'Rivers'
- 合戦 kassen IPA:[kas.sẽɴ] or [kasːẽɴ] 'Battle'
Exist different techniques for the different versions of the software.
As was mentioned before, the Japanese Phonetic system wasn't designed to allow the consonant be reproduced alone, if the user tries to encode it without a vowel this will generate an almost unaudible loop sounding as an electronic buzz. However if the consonant is in middle of two reproducible notes or syllables, the system is capable of hand it better, making possible encode it alone. This permits to use it to extend the some consonant.
For increase the length of a consonant the user must create a gap between the the preceding syllable and the next one containing the consonant to extend. Then fill the gap with a short note containing the consonant phoneme to extend, without a vowel.
It's important that the note preceding the consonant alone must end it vowel, if isn't the case the synthethizer won't be capable of hand it, producing an undesired chop. Also it's important emphatize that although this method allows extend the consonants, the system stills struggles with the consonants encoded alone, specially if these ones are too long. This can generate sound loops or distortion of the phoneme, so it's important not abuse of the method.
For the third version of the software, the parameter Velocity (VEL), was corrected, now effectively affecting the length of the consonants when this one is modified. This, added to the addition of the devoiced phonemes allows effectively modify the length of consonants without utilize complicated techniques or post-edition steps as occurred with VOCALOID2.
|Symbol||Classification||IPA Symbol||Sample Hiragana/ Kunrei-shiki Romaji||Notes||Related Phonemes|
|[a]||vowel||äopen central unrounded vowel||あ a|
|[i]||vowel||iclose front unrounded vowel||い i||
|[M]||vowel||ɯᵝ or ɯ͡βclose back compressed vowel||う u||The japanese "u" is neither rounded [u] nor unrounded [ɯ], but compressed.||
|[e]||vowel||e̞mid front unrounded vowel||え e|
|[o]||vowel||o̞mid back rounded vowel||お o, を|
|[k]||consonant||kvoiceless velar plosive||か ka, く ku, け ke, こ ko||
|[k']||palatalized consonant||kʲ palatalized voiceless velar plosive||き ki, きゃ kya, きゅ kyu, きぇ kye, きょ kyo||Palatalized /k/.||
|[g]||consonant||gvoiced velar plosive||が ga, ぐ gu, げ ge, ご go||
|[g']||palatalized consonant||gʲ||ぎ gi , ぎゃ gya, ぎゅ gyu, ぎぇ gye, ぎょ gyo||Palatalized /g/.||
|[N]||consonant||ŋvelar nasal||が ga, ぐ gu, げ ge, ご go, ん n-n'||Nasalized /g/. Also is an allophone of the /n/ before an velar consonant.||
|[N']||palatalized consonant||ŋʲ||き゜gi , き゜ゃ gya, き゜ゅ gyu, き゜ぇ gye, き゜ょ gyo, ん n-n'||Palatalized nasal /g/.||
|[s]||consonant||svoiceless alveolar sibilant||さ sa, す su, せ se, そ so, すぃ si||
|[S]||palatal consonant||ɕ or ʃʲvoiceless alveolo-palatal sibilant||し shi, しゃ sha, しゅ shu, しぇ she, しょ sho||Palatalized /s/. The X-SAMPA symbol incorrectly suggest it's a /ʃ/, although both phonemes sound similar they aren't the same one.||
|[z]||consonant||zvoiced alveolar sibilant||ず zu, ぜ ze, ぞ zo||Often used between vowels, however not all the Japanese speakers use this sound.||
|[Z]||palatal consonant||ʑ or ʒʲvoiced alveolo palatal sibilant||じゅ ju, じぇ je, じょ jo, じゃ ja, じ ji||Palatalized /z/, often used between vowels, however not all the Japanese speakers use this sound. The X-SAMPA symbol incorrectly suggest it's a /ʒ/, although both phonemes sound similar they aren't the same one.||
[z] (depalatalized) [dZ] (affricated)
|[dz]||consonant||ʣvoiced alveolar affricate||ざ za, ず zu, づ zu, ぜ ze, ぞ zo, じゃja, じ ji, じゅ ju, じぇ je, じょ jo||Often used at the beginning of word or after んn, however some Japanese speakers also use this sound instead of z or Z.||
|[dZ]||palatal consonant||ʥvoiced alveolo-palatal affricative||じ ji, ぢ ji, じゃja, じゅ ju, ぢぇ je, じょ jo||Palatalized /dz/ or /d/, some Japanese speakers use this sound instead of z or Z. The X-SAMPA symbol incorrectly suggest it's a /ʤ/, although both phonemes sound similar they aren't the same one.||
|[t]||consonant||tvoiceless alveolar plosive||た ta, て te, と to, とぅ tu||
|[t']||palatalized consonant||tʲ||てぃ ti, てゅ tyu||Palatalized /t/, usually used into non-Japanese words incorporated to the language.||
|[ts]||consonant||ʦvoiceless alveolar affricate||つ tsu, つぁ tsa, つぃ tsi, つぇ tse, つぉ tso||
|[tS]||palatal consonant||ʨvoiceless alveolo palatal affricate||ち chi, ちゃ cha, ちゅ chu, ちぇ che, ちょ cho||Palatalized /t/. The X-SAMPA symbol incorrectly suggest it's a /ʧ/, although both phonemes sound similar they aren't the same||
|[d]||consonant||dvoiced alveolar plosive||だ da, どぅ du, で de, ど do||
|[d']||consonant||dʲ||でぃ di, でゅ dyu||Palatalized /d/, usually used into non-Japanese words incorporated to the language.||
|[n]||consonant||nalveolar nasal||な na, ぬ nu, ね ne, の no, ん n||This consonant can be articulated without a vowel.||
|[J]||consonant||ɲ or nʲpalatal nasal||に ni, にゃ nya, にゅ nyu, にぇ nye, にょ nyo||Palatalized n, this phoneme also appears as allophone of /n/ before palatal.||
|[h]||consonant||h voiceless glottal fricative||は ha, へ he, ほ ho||
|[h\]||consonant||ɦvoiced glottal fricative||ぁ xa, ぃ xi, ぅ xu, ぇ xe, ぉ xo||Intervowel /h/. Only works for [e] and [o].||
|[C]||palatal consonant||çvoiceless palatal fricative||ひ hi, ひゃ hya, ひゅ hyu, ひぇ hye, ひょ hyo||In the Japanese is perceived as a palatalized h.||[h] (depalatalized)|
|[p\]||consonant||ɸvoiceless bilabial fricative||ふ fu, ふ fwa, ふ fe, ふ fo||
|[p\']||palatalized consonant||ɸʲ||ふぃ fi, ふゃ fya, ふゅ fyu, ふぇ fye, ふょ fyo,||Palatalized /ɸ/.||
|[b]||consonant||bvoiced bilabial plosive||ば ba, ぶ bu, べ be, ぼ bo||
|[b']||palatalized consonant||bʲ||び bi, びゃ bya, びゅ byu, びぇ bye, びょ byo||Palatalized /b/.||
|[p]||consonant||pvoiceless bilabial plosive||ぱ pa, ぷ pu, ぺ pe, ぽ po||
|[p']||palatalized consonant||pʲ||ぴ pi, ぴゃ pya, ぴゅ pyu, ぴぇ pye, ぴょ pyo||Palatalized /p/.||
|[m]||consonant||mbilabial nasal||ま ma, む mu, め me, も mo||Also is allophone of /n/ in front labial consonants. This consonant can be articulated without a vowel.||
|[m']||palatalized consonant||mʲ||み mi, みゃ mya, みゅ myu, みぇ mye, みょ myo||Palatalized /m/.||
|[j]||consonant||jpalatal approximant||や ya, ゆ yu, よ yo, いぇ ye||
|||consonant||ɾalveolar flap||ら ra, る ru, れ re, ろ ro||Although the X-SAMPA suggest that this phoneme is a alveolar tap, technically is an apical postalveolar flap undefined for laterality, hence the Japanese /r/ tends to sound somewhat between a ɽ and a ɺ. If the consonant has a more R-like or L-like sound, depends of its context.||[4'] (palatalized)|
|[4']||palatalized consonant||ɾʲ||り ri, りゃ rya, りゅ ryu, りょ ryo|| (depalatalized)|
|[w]||consonant||w͍ or wᵝcompressed labio-velar approximant||わ wa, うぃ wi, うぇ we, うぉ wo||Similar to its /u/, the Japanese /w/ is compressed.||
|[N\]||consonant||ɴuvular nasal||ん n||/n/ at the of end of word.||[n]|
- The Japanese Phonetic System actually uses the symbol <¥> instead of <\>. However, for easier comparison with their X-SAMPA and to synchronize with most of the keyboard, typing <\> will be input as <¥> in the sythesizer, the wikia will prefer this notation among the articles.
- Crypton’s VOCALOIDs, including KAITO and MEIKO, have almost the same Japanese phonetic system. To use [z], [Z], [h\], [N] and [N'] , users need to edit the phonemes, not entering kana-characters.
- Kagamine Rin/Len Act 1 can pronounce [h\] while their Act 2 cannot (comparison of consonant sounds Act 1, Act 2).
- Vocaloids of Internet Co., Ltd., such as Gackpoid or Megpoid, mostly share the same system as Crypton's, but they do not have [z] and [Z] sounds. As is often the case with the Japanese language, they are replaced by [dz] and [dZ].
- Japanese VOCALOID2 voicebanks can combine a and i phonemes (eg. [w a i]) but not with the original VOCALOID voicebanks. The workaround is to simply use the y consonant. (eg. [w a j])
- [N\], [N] or [n] alone tends to be pronounced as "ng". This is the basis for Japanese vocaloids being used for South-East Asian languages.
- [N'] followed by a vowel different to [i] may produce odd results, however, due to its use within the Japanese language there is no actual call for this phonetic to be followed by a vowel different to [i].
The Japanese language is by far the most popular VOCALOID language and most well refined as it is the language that create the cultural phenomenon that made VOCALOID popular in late 2007, though this popularity has levelled off since 2014. Japanese vocals make up the largest selection of the VOCALOID vocals available for purchase as a result.
It is much quicker to produce a Japanese vocal and much cheaper in addition and overall between popularity of version and cheapness/speed Japanese VOCALOID have been produced faster, have more content in their releases and overall cheaper then all the other languages combined.
- Conversion Lists
- Interwiki articles
- ↑ link
- ↑ link
- ↑ explanation for accents in singing and also a lack of
- ↑ [ http://www.todayifoundout.com/index.php/2013/08/why-british-singers-lose-their-accent-when-singing/ "Why do British singers lose their accents?"]
- ↑ https://weibo.com/5146173015/Gckc5ax0y
- ↑ 
- ↑ http://a0010.web.fc2.com/text/v3memo/index.html
- ↑ http://ahou2chome.sakuratan.com/misc/mikuvoice/lesson02.html
- ↑ Japanese Phonetic System of VOCALOID KAITO
- ↑ Japanese Phonetic System of Megpoid
- ↑ Japanese Phonetic System of Gackpoid