In 2017, Hatsune Miku’s 10th anniversary was approaching. The second version of her English phoneme voicebank – titled “V4 English” – was newly released, and I decided to look at a demo song out of curiosity. I didn’t really follow the scene directly or listen to a whole ton of Vocaloid music; my music preferences usually follow specific artists I like (where some happened to focus on Vocaloid) rather than dedicating myself to a genre. In any case, the comments section gave me a bit of shock.
Essentially, the majority of the comments for the demo tracks were about how bad she sounded. Was it an improvement over the first version? According to them, a little bit… but not really.
I didn’t agree with this at all, and decided to spend 2017 by throwing my hat in the ring and buying her software package and seeing what I could do with it. Theoretically, I thought, if the phonemes are correct, then it’s a matter to applying the correct linguistics theory to string together the words. All of my projects that year were about making an album’s worth of songs in that regard.
This article is about one part of that adventure. There are two big ‘stories’ or ‘retrospectives’ I’d like to share, which I consider to be the two biggest things I studied: One is about developing a proper linguistic understanding of Miku to make her sound truly fluent in English. The other is about one of the major hurdles of translating and developing a pre-existing song, which is writing good lyrics. Overall, it feels like I had to combine a lot of different disciplines to invent a Vocaloid song that I really liked on a personal level: Linguistics, music theory, Japanese subtext, and English poetry were all things I had to brush up on.
To simplify the article, I’ll focus on only one song, and you can listen to the focus of this article, here.
That said, since I’m writing this retrospective six years later on Miku’s 16th birthday… well, hopefully I remember everything I did. An article focusing on the linguistic theory I developed for Miku may come another time.
About English Covers
I’ve tried listening to English covers several times over the years but only managed to like a small handful of songs. By ‘English Cover’, I refer to Japanese pop songs that have been fan-translated into English lyrics, and sung on YouTube. It’s a fairly popular thing to do, but I never managed to seriously like it as a genre to dig deeply into.
The reason for this is that it, quite simply, doesn’t sound right to my ears. Japanese songwriting is fundamentally different from English songwriting, and even when it’s a song I’m not familiar with in Japanese, there’s usually something that sounds very off with the way it’s being sung, regardless of skill.
When doing a project I like to study several birds with one stone, so to speak, so I decided to also translate and localise one of Miku’s first, definitive songs: ‘Hajimete no Oto’ by malo. This meant that I had to think about why I didn’t like English cover songs, and write in a way that addressed the pitfalls. In the end, I decided to label these as English adaptations.
I’m going to layout this article as two distinct sections:
First is general theory, and then I’ll break down my favourite song and talk about my lyrics.
There’s also a Closing section after that.
Translation and Music Theory
So how is Japanese songwriting different from English songwriting? Studying this and listing out these things was rather important before thinking of a new set of lyrics. Here are all the concepts I remember applying. Some of these are differences in poetry and music theory, and some relate to translation styles.
1. Case Studies
I recall studying three major case studies to figure out how Japanese differed from English, all from professionals in the translation or music industry.
The most significant one is Utada Hikaru’s ‘Hikari‘ and ‘Simple and Clean‘. All of the theme songs for the Kingdom Hearts series have dual-language versions. I don’t know where she said it, but I’m fairly certain that during the development of this song (or an interview for KH2’s Passion/Sanctuary) she said that writing these theme songs were fairly difficult, due to the differences in music theory. In the end, ‘Hikari’ and ‘Simple and Clean’ ended up having very different main melodies, despite having the same backing track.
The second case study is Godiego’s Japanese and English versions of ‘Gandhara’, written in 1978. This one is interesting to study because it’s sung very slowly, which makes it easier to study; it also doesn’t turn out to be a 100% literal translation, though it matches very well. They also somehow managed to have a very similar word density in both versions, even though there’s the syllable problem.
The third case study is the Australia-exclusive version of Cardcaptor Sakura and its translated opening songs. Yes, Nelvana really did simultaneously produce two versions of the anime dub with redirected acting, one for North America and one for export. This one is least interesting since the lyrics mostly do not match, but is still worth looking at to see the cadence while still adding lyrical shoutouts to the original song.
2. Notes and syllables
The most significant difference I found in Japanese music compared to English is that each and every note in a song is associated with only one mora, or syllable. As a syllabic language, this makes Japanese music slightly easier to sing, and is probably neccesary to even make the song comprehensible. This can also apply to glottal stops such as ‘-tte’, but can be skipped in specific types of words due to the way the modern Japanese accent works.
(In essence, some words ending in ‘-shite’ or ‘-su’ can have a sound drop, so it sounds more like a single mora of ‘-shte’ or ‘-ss’. This is exclusive to modern spoken Japanese and is not reflected in writing, so the sound can return if you speak slowly or enunciate every syllable, so it appears in Japanese music semi-often.)
On the other hand, English lyrics can and can occasionally stretch a single syllable over several notes for emphasis. Not taking this into consideration can increase the verbosity of English lines, but it’s difficult to combat, due to the higher number of melodic notes that seem to be in Japanese music in general.
3. Note splitting & Note merging
The syllabic density of a Japanese word can also be higher than the same English word. While written Japanese is about 2.5x more information dense than English on a per-character level, it is actually reversed in spoken form.
If you listen to some music such as Gandhara, you should be able to hear a number of semiquavers in the Japanese version actually sung as quavers, thereby reducing the number of syllables required in the English version. This largely only works on bars that have the same note, as this retains the same overall melody. You can’t do as much if the song is highly melodic with lots of changing notes, but it is still possible, if you make subtle changes to the note lengths.
This can also work in the opposite direction, but can make the song sound like it’s being sung too fast in English. Not figuring out a way around this and making a really verbose English cover is probably what throws me off the most with these songs, but again, there’s a limit to how much you can do unless you really work on merging notes and make the English sing more slowly.
4. Rhyme and Subject-Verb order
Basically, Japanese doesn’t have rhyming as a concept, and instead can rely more on making variations of a phrase. If you’re not paying attention or can’t understand Japanese, you can sort of audibly hear something that sounds like a rhyme on occasion, but it’s actually reusing the same verb or conjunction. It’s not perfect, but taking English poetry into consideration is worth trying.
The reason for this is likely because Japanese is a language with a sentence structure of subject-object-verb, while English is a subject-verb-object language. Since Japanese sentences end with a verb, the rhyme-like cadence is achieved by repeating the same sentence in a different way.
5. Woolseyisms & Instinct
The basic concept of a Woolseyism (or at least my interpretation of it) is that if you can come up with a completely different line that happens to perfectly fit the character profile and emotional tone of the original line, you should use it. It can be far more memorable and natural to an English speaker.
I’m not sure if there’s a good way to describe what skills you want to have to try to pull one of these off, but I would say that your level of fluency in the target language (ie. English) needs to be very high. However, being just a native speaker is a bit of a trap and is actually not enough; what you really need is a high level of literacy, and maybe just a good feel for the language.
6. Synonyms and half-synonyms
I don’t think ‘half-synonym’ is actually a thing, but this is where translation can really be more of an art. I’m just making up the term since I don’t know how to describe it.
Sometimes there’s just no way around a wonky line and the only thing left to do is find a related word that thematically fits, but isn’t a strict synonym you find listed in the thesaurus. This goes into the idea of translating the intent of the words, rather than the literal meaning. Since we’re talking about music, I think it’s okay to do some poetic license with, well, poetry.
7. Cadence and beats
Another added difficulty is that ideally, in a 4/4 song, you want the words you want emphasised to be on either the first or third beat of a bar. There’s not much to explain for this one since this is essentially the important part of poetry and rhythm.
In the end, some of these ideas made me decide to label it as an adaptation, because I’m taking more effort to adapt the song – I even slightly modified the melody and timing to fit the words.
Verse-by-Verse Breakdown
In this section, I’ll essentially be going over certain lines and sharing notes on my thought process.
The song in question is ‘hajimete no oto’ by malo, and was one of the first songs written with Miku that framed her as a character of sorts. It ended up being one of the defining influences of her popularity, and I thought it was a really appropriate choice to challenge myself on.
Also, no one had ever made an English cover of this song before, which I thought was a shame.
The song itself is a ballad, and creates a personality for Hatsune Miku. The general concept embraces the fact that Miku is not a real girl, but a piece of singing software that requires input; she wonders about simple, almost infantile concepts like one’s first sound and words, but she also doesn’t know anything about the real world: She can’t feel or sense anything, and time has no meaning. She is only capable of taking your input and making a song with it, and it will be the same every time.
Anyway, let’s reminisce over my version of the lyrics.
| Hajimete no Oto (by malo) | My First Melody (by Eiri) |
|---|---|
| What was the first sound? Your first sound… This one is mine, So right now, I’m happy. | My very first melody was a simple little thing, I wonder what you had for your first song? I’ll share this one with you, and I’m happy to sing it all for you, and everyone here. |
This was really the most difficult part of the song I remember struggling with, and it’s that ‘hajimete no oto’ is a seven-syllable phrase, and ‘first sound’, the literal translation, is only two. The fact that the title of the song is actually in the lyrics makes it a little harder. In the end, I couldn’t find a good way around it while reducing the syllable count, so I went with ‘My very first melody’.
Changing the start of the song and title to be about ‘melody’ instead of ‘sound’ was really hard to accept, but the overarching idea I ended up following was that I would at least retain the ‘story’ in the ballad, even though I’m adding filler in-between the beats. To retain the original meaning of the song, I moved the verse about ‘sound’ to verse 2, which is about the first word.
In lines 3-4, to increase the word count you can see I flipped the perspective while retaining the meaning.
| What was the first word? Your first word… I can’t speak, So I sing, like this. | My very first words were a simple little sound, I wonder what you had for your first sound? I don’t know how to talk at all, and I never will ’cause all I know, is how to sing a song |
You can see here that I essentially merged the original Verse 1 and 2 into a single verse about sounds and words, just so I can cover the syllable count. I also end up adding more emphasis in line 3, which is something that’s from a later verse.
In line 4, you should be able to hear ‘a song’ stretch over several notes. Another thing to consider with the English lines is that you want the words you want stressed to end up on the correct beats. That’s why ‘Because’ is truncated to “’cause”, because otherwise the word ‘know’ wouldn’t be stressed on the first note!
If bold is the first beat and italics the the third, the stresses went like this:
I don’t know how to talk at all, and I never will
’cause all I know, is how to sing a song
| In time, the days go by, and years go by, And the world will fade away. But as long as I have your light, I will always sing. | And I know I’ll never see what all of it means, not even when the world fades away. But I’ll remember you, and all the words you gave to me and I will sing – I’ll sing into the night forever. |
Line 1 here is, admittedly, a pure Woolseyism. The Japanese lyrics are ‘yagate hi ga sugi, toshi ga sugi‘, and despite the different meaning, the English line somehow came to me very quickly, with the added bonus that it rhymed with the original Japanese.
The only issue here is that this line does match the meaning in the next verse, but bringing the story of the ballad forward much like moving the ‘sound’ line slightly backward, felt worth it.
Lines 3 and 4 is, oddly enough, another reversal. The ‘light’ metaphor probably refers to user input, so I made it literal, and then I still made use of the light reference by extending Line 4, by implying that your words gave her a light in the darkness.
| The colour of the sky, the scent of the wind The depth of the ocean, and your voice – I don’t know them, but a song— I can raise a voice, and simply sing a song. | Soaring in the sky, diving into the sea, Wandering the land until I hear your voices to me, I’ll never do all of those things, I will never feel them, but I know that I can always sing a song, and then |
Instead of Woolseyism, this is more like poetic liberty, and took a lot of work to figure out. The key thing to realise in this verse is that Lines 1 and 2 are listing out things that Miku is physically incapable of sensing.
In my version, I made it to be about what Miku is incapable of doing, but still went for actions that have strong sensory experiences. In addition, there’s a bit of redundancy with the sky and wind, so I made the three experiences be about the land, sea, and sky – essentially improving the variety of the message and making it more evocative, I think.
You can also see that 1/2 and 3/4 rhyme, and Line 1 has lots of alliteration.
| If something can reach you, I’ll do it again, any number of times, and It’ll be the same, every time, Just like the first sound. | If you ever reach out, I will be there Singing for you, and singing forevermore I will sound as you say, any time and any day and I’ll be there, for that first melody. |
Line 1 is another reversal for some reason… It seems like I essentially changed all the passive lines into more active ones that address the listener.
Line 3 honestly felt like a ‘genius’ moment to me when I came up with it since not only is it a very poetic interpretation of the line, but it’s a direct reference to the original song’s name about ‘sound’, and it rhymes!
Not only that, but it’s a multi-layered meaning where I am both using the verb ‘sound’, and and using the noun ‘sound’ as a verb – because Miku’s only ability is to generate ‘sound’ as a noun, and not speak like a human.
In other words, “I will sound as you say” can either be interpreted as literally sounding as you intend like a program – or, it can refer to how she literally only has the ability to make sounds, while you have the ability to speak. I added a throwback to Verse 2 to bring the song full circle!
| Are you familiar with that first sound? Your first sound… Anywhere in the world, I will sing Each and every first sound. | My very first melody was a simple little thing, but I wonder if you know your own song? No matter wherever you are, no matter where you choose to go, I will be here for everyone, and their song. |
I think this was the one part that I wasn’t quite satisfied with. That’s because it’s an echo of the first verse, and that was also a real mess to work with. Lines 3 and 4 are exceptionally verbose in my English version. I didn’t even find a proper rhyme for lines 2/4.
That said, just looking at Line 3 shows just how bad it can get, if you don’t figure out a way to reduce the number of notes in a Japanese song. Since English words have fewer syllables overall compared to Japanese, you constantly need to figure out how to add filler, which can have a serious effect on the poetic strength of the English lyrics.
Closing
Overall, I suppose this is one of those cases where it’s a project I’m personally proud of, but writing about it feels like I’m patting myself on the back a little much, isn’t it? There were a lot of things I had to be clever about or study hard on, and figuring out how to write lyrics that felt really good to me, because I didn’t feel satisfied with the efforts I’d seen so far, is a bit of a strange motivation.
However, I think there’s something to be said when I caught myself singing my own translation to myself instead of the original lyrics, six years later and not fully remembering what I did. Some lines just really stuck in my head after all that time, and I wish I could have originally shared it with more people.
But I’m not sure who can appreciate it. I’ve actually wanted to write this for years, and did try once, before I took it down. The problem I ran into though, is something that I think I still ran into when I became a Vtuber, a year later. I was making these songs and exploring something I thought was very interesting, but I honestly had no idea who to share these ideas with. I sorta still don’t.
So, I write this now, since I’m reminiscing a bit.
There is also a second potential article I can do, concentrating on the linguistic theory I developed to make Miku sound truly fluent without aid from subtitles, but that is for another time.
Eiri Sanada 