I want to make a Vocaloid sing in English! I tried writing English lyrics.

favorite_borderLast updated:2024/8/4

• DTMermilk cocoa
  Former IT engineer. I use a wind synthesizer. I make songs using Vocaloid and CeVIO, and sometimes I invite a vocalist to create songs as well. I originally was a half-baked DTM user who only made backing tracks for instrument practice, but after discovering Vocaloid and following its evolution, I fell deep into it—just a music lover. I use Vocaloid 2 Megpoid, Vocaloid 3 IA, and ONE from CeVIO Creative Studio.

Trying English lyrics with a Vocaloid!

This time, I’d like to get a bit technical and actually try out how to make a Vocaloid sing English lyrics.

Here’s the piece I tried.

https://www.

youtube.

com/watch?v=fSXsUC4O3FI

It's “We're All Alone” by Boz Scaggs.

Vocaloid isIAI tried making it sing using it.

I think there are techniques that can be used outside of English lyrics as well, so please stay with me until the end.

I want them to sing in English! But there’s no English library!

Except for libraries that originally advertise English support—such as Hatsune Miku, Megurine Luka, and Megpoid—there is no English pronunciation data, so even when they attempt English, the words are pronounced with Japanese intonation.

This issue is a major obstacle to aiming for English-like singing.

For example, even for a single word like “true,” it’s not as simple as just typing “to,” “ru,” and a long vowel mark.If you don’t make an effort to pronounce it with the accent centered on the “ru,” it’ll end up sounding blatantly like clichéd Japanese-style English.

To make this even a little more convincing, we’ll need to get creative with how we tweak the parameters.

With VOCALOID4’s cross-synthesis, it’s become easier to achieve a more English-like sound.

Since VOCALOID4, a feature called Cross-Synthesis has been added.Previously introduced as well.I have done it before.

This feature lets you keep two VOCALOID libraries on standby so you can either mix them by adjusting parameters or switch to the voice that was on standby.

You can use it much like adjusting the mix amount of an effect inserted directly on a channel.

Reference:VOCALOID4 Announced! Highlights include growl and cross-synthesis that morphs between singing voices: Ken Fujimoto’s DTM Station

By leveraging this, it becomes easier to produce expressions that were difficult with Vocaloid, such as rolled-r sound-like pronunciations and emphasizing only the accented part within a word.

In this song, I have them sing with slightly rolled pronunciations in words like “more” and “close,” and since there are many parts where “be” is sung strongly, I incorporate cross-synthesis throughout.

One word per note is the norm.

One problem you’re bound to run into when dealing with English is that, without mercy, one word is assigned to a single note of the score.

You’re still better off if the sixteenth note has a pronunciation that becomes three characters in Japanese, like “cry” or “out,” but it’s not uncommon to find words of four characters or more, such as “chance” or “wait.”

Vocaloid isMIDIWe have a base input system.

I understand that this offers the advantage of easy integration with DAWs and other systems in terms of compatibility, but for the reasons mentioned above, I actually don’t think it’s well-suited for entering lyrics.

In cases like thisIn Vocaloid, for pronunciations of three or four characters, a single note must be divided into as many segments as there are characters.

This is the starting lyric word, “out.”

We’re dividing a sixteenth note into three parts so that we can pronounce “out” within the duration of a single sixteenth note.

In cases like this, turn off the DAW’s snap feature and ...If the cut tool isn’t set to allow cutting anywhere, workflow efficiency will deteriorate significantly, so turn off the snap feature when entering English lyrics.I'll leave it that way.

In the song I tackled this time, besides words like “out” and “cry,” there was also a phrase starting with “close,” and I struggled a bit because it was assigned to very short notes.

Sounds that are produced and sounds that disappear

This alone still doesn't convey much of an English-like quality.

For example, in the case of “out,” it’s written as three katakana characters, アウト, in the Japanese lyrics, but in actual singing the final ‘t’ is almost never pronounced.

That said, typing “a-u” doesn’t yield the value of three mora; it just stretches the “u,” and the absence of the supposedly inaudible “t” creates an unnatural break in the flow before and after it.

I'll make a few tweaks to help it sound more natural when sung.

When you right-click after selecting a note in the Vocaloid editor, a menu opens.

If you select “Note Properties” within it, detailed statuses of the note will be displayed.

Here you can configure the note lyrics, accent intensity, how to handle note values (decay), and whether to connect smoothly to the next note (portamento).

Each one has significant meaning, but the most important part in the “to” where the “out” is not pronounced is the phonetic (pronunciation) symbols circled in red.

The pronunciation of the Vocaloid is based on this phonetic notation.

Even if the lyric is entered as “a,” if the phoneme is set to “to,” the pronunciation will be “to.”

And when you add “_0 (underscore and zero)” after this phonetic symbol, this soundvoiceless soundis treated as

This is a command that makes you produce a sound like air smoothly escaping.

In English, there are many words where sounds like these drop off at the end.

In this case, the “out” is followed by the lyrics “outside the ……”.

Written in Japanese, it becomes “アウト サイド ザ,” creating a connection between the syllables ト and サ.

In actual singing, the 'to' and 'sa' are voiced almost seamlessly, so the 'to' isn’t audible.

To get closer to this kind of pronunciation, I’m making the “to” sound voiceless so it effectively disappears, which helps the singing sound more English-like.

There are also words whose sounds disappear even at the end of a phrase (when nothing follows).

It’s the word “alone” in the part where they sing “we’re all alone,” which is also the song’s title.

In Japanese it would be “arōn” (alone), but even when I listen to the lyrics, the “n” isn’t clearly audible, and if we insist on calling it “arōn,” the sense of rhythm suddenly gets worse.

In most cases, when a phrase ends with the sound “n,” you don’t need to add any characters at the end.

The final “n” isn’t included, but by lowering the dynamics at the tail of the elongated “ro” and letting it be swallowed, we were able to express the unvoiced “n.”

Was 'Let it go (rerigō)' correct?

FrozenThe famous lyrics “Let it go.”

It was sometimes represented in Japanese pronunciation as “rerigō.”

Grammatically, it should be “it” in English, but in the Vocaloid world, “rerigō” would actually be correct.

I only typed “ui o ra ro” for “we’re all alone.”

The “are” in “we’re” is crammed into a sixteenth note and is almost inaudible, while the following “all” is more important. Also, the “ru” sound at the end of “ōru” (Japanese for “all”) almost connects to the “alone” that follows, so I let it collapse into an approximate “ra” by blending “ru” + “a.”

It brings the phrase together by greatly lowering the accent on “ra,” removing the sense of attack and allowing the “ra” sound to blend in smoothly.

The phrase “let it out” that appears in the latter half of the chorus.

let it all begin.

The lyrics are achieved with exactly the same input as “Let It Go.”

If you crank the accent and decay all the way up on the initial “re” and make the “ri” a slightly weaker accent, it strangely starts to sound like “let it.”

This time I deliberately went for a “Let It Go”-style feel, but you can get a similar vibe with “ti” or “i” as well.

I think this is the fun part where you can try various approaches depending on the library and the vibe of the singing style.

we’re all alone feat.

IA&IA Rocks

Let's listen to the result of having it sing with all the tuning we've done so far.

https://www.

youtube.

com/watch?v=Dh-9WyKS8C0

What do you think?

There are still plenty of challenges, but it’s started to sing in a pretty English-sounding way.

With a bit of ingenuity, a Vocaloid can handle even difficult phrases, so try having it sing a variety of songs.

I tried it with CeVIO too.

Last time "Winter RivieraI tried having CeVIO sing it too, using the same song.

This time, the one singing is ONE, who is released by the same label as IA, 1st Place, and is referred to as IA’s younger sister.

Please listen and compare to hear what’s different.

https://www.

youtube.

com/watch?v=kutxPxt94n4

I’ve introduced vocal tuning in CeVIO separately, so I’d be grateful if you could take a look at that as well.

Reference:ABC sung with CeVIO Part 1

Reference:ABC Part 2 Sung with CeVIO

Have you gotten a general sense of the characteristics of both Vocaloid and CeVIO?

Everyone has strengths and weaknesses, but with some ingenuity, many things are possible.

This time I introduced just a small part of those ideas, but I hope it will serve as a hint to someone.

Please try having the Vocaloids sing your wonderful songs too!