Idoru

by Belle Waring on December 15, 2013

Sooooo, the youngs. you may have heard they like Justin Beiber or Rhianna or something. They don’t. They like computer constructs, only one of which is human, and we hear her voice only, and anyway there is some debate about whether she’s canon. Vocaloids! The original technology was invented by Kenmochi Hideki at the Pompeu Fabra University in Barcelona, Spain in 2000. Backed by the Yamaha Corporation, it developed the software into the commercial product “Vocaloid.” (ボーカロイド Bōkaroido). (This product exists separately from the Vocaloids I’m talking about and is used to generate back-up vocals and other things like that in ordinary pop songs). The most popular is naturally 01, Hatsune Miku. You can even see her perform live! (You should really watch this—it’s not clear quite how bizarre the scene is till partway through.) Her ‘voice’ is compressed into the upper range of human hearing, and beyond what any human could sing. But it’s not merely a person’s voice sped up; it’s constructed (though some samples were taken from a Japanese actress).

If you are a fan of Hatsune Miku, you can buy her voice software. Then, you can write a song, and feed it to her, and she will sing it! And if you have a friend who draws manga-style drawings well, she can draw drawings for you, and then the “camera” so to speak can move over the drawings during the course of the song. Full animation is expensive; no one but the owners of Miku (Crypton Future Media) can afford it, I think, unless there are some really rich Miku fans, which there probably are, and that’s how you end up with non-canon Miku, which is crazy, because she’s Vocaloid 01! There are also manga, an alternate universe Hachune Miku, a variety of games, actual albums released on Japanese label Exit Tunes that charted at #1, a billion games, and some anime (but I’ve never seen any of these because I just rock it from YouTube, where videos get posted after they get reviews and feedback on Japanese video-posting site Nico Nico Douga.) Perhaps most awesome, the Japanese Venus space probe Akatsuki has three plates depicting Miku on it. Not sure whether she’s carrying leeks in any of them, but she should be, because leeks are her thing. She’s, just, always got leeks. I don’t know.

Vocaloids 02 and 03 are Rin and Len Kagamine, who are blond twins, one girl and one boy respectively, now with 75% less Blood of the Walsungs. OK, 70%. 04 is…some dude. Our daughters have an aversion to the synthesized male voices (which appear to be possibly—no, pretty much we know this, they are sampled, but by artists who don’t wish to reveal their names because then people will be able to ‘force’ them to sing anything they want.) 05 is Luka, she has some good songs, 06 is Gumi, then…something…Lily’s OK…etc. 99 is a human woman’s voice, but we don’t know her name and never see her real face. Overall, moving from 01-99 the voices become more like human voices, lower, more—I don’t know, irregularities? Roughness?

The songs (and videos) are sometimes funny, sometimes super-depressing, and sometimes creepy (there is a problem with whether you can use Rin and Len, say, to make child porn songs.) Obviously they are often about how it’s alright to be imperfect, or how difficult it is to make connections with real people, or…what it’s like to be an AI whose master cruelly re-programs it once a day with an unattainable goal. Funny stuff!

Is there any difference between life and death? Between happiness and its absence? Luka says no. You should go all the way through with this one as the music’s unobjectionable and it has the most ironic lyrics to program a Vocaloid to sing for you ever:

This is Gumi, who suffers from anxieties peculiar to a modern city-dweller? Or not:

I freely acknowledge that this is a poorly animated manga CGI girl who sounds like you have caught a bunch of bees in a Mason jar and are just shaking them up and down BUT it is important to watch/listen to the song enough (reading the subtitles) to realize that this happy bunny funtime blonde robot girl Rin Kagamine wants to grab you by your slender pale throat so that you may dive, together, into the core of a fusion reactor, where everything will become white and all bad memories will be erased. See? Now she’s like a billion times creepier, right?

Does Rin ever rock out though? Yuh huh, just as much as Green Day does.

{ 22 comments }

1

oldster 12.15.13 at 2:59 pm

“Not sure whether she’s carrying leeks in any of them, but she should be, because leeks are her thing. She’s, just, always got leeks. I don’t know.”

Cause she’s Welsh, obvs. But you can hear that in her accent, anyhow.

2

Belle Waring 12.15.13 at 3:06 pm

Oh, yeah, duh. Do I ever feel dumb now.

3

oldster 12.15.13 at 3:24 pm

Despite what Catherine Zeta-Jones might lead you to believe, Welsh girls can be animated.

4

Consumatopia 12.15.13 at 6:46 pm

This is good. I was getting kind of depressed that the future wasn’t displacing humanity fast enough. Pretty soon everyone will be able to “force” everyone to sing.

Someone has to link PuppyCat, right? http://www.youtube.com/watch?v=muIisW6L1Sk

5

SusanC 12.15.13 at 7:14 pm

I’m a big fan of Vocaloid – I have the software on my computer (plus Cubase, SampleTank, etc., to make the whole music synthesis system). I prefer Megurine Luka’s voice, but it really depends on the genre of song .. Hatsune Miku might sound better if you’re doing a J-Pop cover.

The thing with the leek derives from a Hatsune Miku cover of the Levan Polka:

http://www.youtube.com/watch?v=kbbA9BhCTko

6

SusanC 12.15.13 at 7:31 pm

Oh, and for low-budget animation there’s always “mechanima” — video game software repurposed to make computer generated animation. If I recall correctly, there’s already a Second Life avatar model for Hatsune Miku, for example.

(I’m suddenly wondering if it’s possible to use something like Microsoft’s Kinect for motion capture as you or your friend dance to the song, and then feed the motion capture into rendering software lifted from Unreal Engine or the Second Life client, using the Hatsune Miku 3D model)

7

fencedude 12.15.13 at 8:42 pm

My favorite Vocaloid song is Magnet, a duet between Hatsune Miku and Megurine Luka which is unapologetically a lesbian love anthem.

As for animations, there is a 3d animation software called “Miku Miku Dance” which…does pretty much exactly what it sounds like. Its actually really versatile and there are tons of user created mods and whatnot for it.

8

John Emerson 12.15.13 at 11:28 pm

Animated indeed”>: The perceived barbarism of the Welsh people was highly gendered and sexualized. As evidence of this immorality they declared that “EIGHT out of every TEN of the women,above the age of sixteen, UNCHASTE and INSENSIBLE to female virtue.” TheCommissioners identified the women of Wales as the primary cause of Welsh immoralityand deviance; their behavior was the most visible signpost of the need for Wales to becivilized by the English.

http://spruce.flint.umich.edu/~ellisjs/Kreider.PDF

9

oldster 12.16.13 at 12:22 am

Thank you, John! That’s very informative.

The Commissioners were all elite Anglican Englishmen educated at
Oxford or Cambridge. None spoke Welsh. They were given letters of introduction to
members of the Anglican clergy in Wales, and 80 percent of the witnesses providing
evidence were also elite Anglicans.
The Reverend James Denning informed the Commissioners
that, “. . . from my experience of Ireland. . . there is a very great similarity between the lower orders of Welsh and Irish – both are dirty, indolent, bigoted and contented.”

The unbridled insolence of the man! Can you believe he called the Irish “contented”?

10

oldster 12.16.13 at 12:25 am

The whole thing is comedy gold.

Commissioner Lingen argued that:
Poetical and enthusiastic warmth of religious feeling, careful attendance upon
religious services, zealous interest in religious knowledge, the comparative
absence of crime, are found side by side with the most unreasoning
prejudices or impulses; an utter want of method in thinking and acting; and
(what is far worse), with a wide-spread disregard of temperance, whenever
there are the means of excess, of chastity, of veracity, and of fair dealing.

The solution, I should thing, is to prevent them from getting access to the means of chastity, veracity, and fair dealing. That will keep them from disregarding temperance!

Or another solution might be to put “whenever there are the means of excess” inside parentheses.

11

Belle Waring 12.16.13 at 2:00 am

SusanC: that’s interesting; we were sort of wondering if the software’s super-expensive, and if the editor’s easy to use. I think that you must program the straight phonemes in (I can’t imagine how else it can work…unless it has a good text recognition software you can correct?) but how do you tell it which syllables to stress in a given sung line? Our daughter would want Mandarin/English/Japanese, that could be difficult, and although there’s probably a Japanese bilingual edition I think it’d have all its controls in Japanese and be a pain. I agree Luka is good. Miku gets on my nerves sometimes though of course there are good songs also…I guess Hocus Pocus must be a Miku song since it’s a flashback to hypothetical abused child Miku. Blue hair ribbons only-style.

I’m kind of surprised there are so few topical comments on the realized cyberpunk awesome. Are CT readers all, like, ‘oh yeah Vocaloids, way to hit us up with some shit from 2010, Belle’?

12

gwern 12.16.13 at 4:45 am

> that’s interesting; we were sort of wondering if the software’s super-expensive, and if the editor’s easy to use.

It’s easily pirated, if that’s a problem.

> but how do you tell it which syllables to stress in a given sung line?

No idea, but I know it’s doable because there’s no way you could get some of the better songs without a lot of control over stress and breath.

> Our daughter would want Mandarin/English/Japanese, that could be difficult

The new Mandarin voicebank seems decent enough, but there is not and probably never will be a single voice which does Mandarin *and* English *and* Japanese well… The quality depends on the voice donor – why does Luka sound great in Japanese and pretty decent in English? Because the original singer is fairly fluent in English. Why does even the new English Miku voicebank sound like crap? Because the Miku singer isn’t.

> Miku gets on my nerves sometimes though of course there are good songs also

I think it’s actually kind of a pity that Vocaloid is so dominated by Miku. I know the Japanese love the chipmunk voice, but they’re a bit of an outlier in pitch preference: most countries would be much happier if more of NND was done using Ia or Luka or someone else.

> I’m kind of surprised there are so few topical comments on the realized cyberpunk awesome.

To be fair, William Gibson got there with the ‘singing Japanese software idol’ way back in… 1996? And of course, Japan did that a few years before in _Macross Plus_ with Sharon Apple.

13

garymar 12.16.13 at 6:01 am

After listening to the youtube link, I read up on AKB48 in some detail, because Belle’s interest may be in the Gibsonian cyberpunkish Idoru aspects of this, but I see it as one fork of a bifurcation in the Japanese Idoru marketing phenomenon: on one side, the “etheralization” of the Hatsune software, versus the “concretization” (Hegelian scholars, help me out here) of the AKB48 girl group, a group of real live girls who become “idols you can meet!”

‘The group’s style is loosely characterized as “bubblegum pop and synchronised dancing” that appeals to pre-pubescent girls as well as older males who support the group’s merchandise.’ Support the group’s merchandise by all means, you older males!

I speak Japanese pretty well, but I can’t make out a single word that Ms Hatsune is singing. That might have just been the Youtube sound quality. The name they picked for her is quite allusive: hatsune means “first sound” or “first bird song” – it’s even a chapter title in the Tale of Genji.

14

gmoke 12.16.13 at 6:21 am

Japanese music industry has a history of manufacturing pop idols on an assembly line basis and then abandoning them, even without the Idoru or anime form factor. Wonder if anyone’s read Spinrad’s Little Heroes which could serve as the flesh and blood model for manufactured personalities with techno enhancements.

15

Neville Morley 12.16.13 at 9:58 am

“Are CT readers all, like, ‘oh yeah Vocaloids, way to hit us up with some shit from 2010, Belle’?”

Well, yes. It’s obvious that Katy Perry, Justin Bieber and all the X-Factor crowd are digital constructs; these Vocaloids have the undeniable charm of retro futurism, the past’s idea of what the future was going to be like, like listening to old Kraftwerk albums.

16

SusanC 12.16.13 at 10:04 am

Belle: The way the software works is that you start with a MIDI file for the melody, assign phonemes to notes, and then adjust the stress/emphasis – there are graphs of stress against time that you can adjust. Adjusting the stress is the tricky bit. Which phonemes are available depends on the actor whose voice was sampled. Luka Megurine is based on Yuu Asakawa’s voice, and as she’s bilingual Japanese/English, the corresponding Vocaloid is OK in both languages (and sort of OK in any language, as long as it doesn’t use phonemes that don’t occur in English or Japanese).

17

Belle Waring 12.16.13 at 1:53 pm

SusanC, gwern: that makes sense regarding the voicebank being usable if the sampled voice has fluency in the various languages. I didn’t think they were so heavily dependent on a single donor voice. The girls would ditch Mandarin because they don’t like any Mando-pop styles particularly. MeiMei drums in actual real life, and well, so if they make songs they can at least have the interest of played keyboard and real drums plus Megarine Luka. SusanC, in the song at the top they have adjusted the stress quite carefully for the final questions, “is it ok?” “it’s ok.” They’re the same set of phonemes in Japanese, it is clear on listening. fencedude: your comment was stuck in moderation for no reason (well a reason-ish, has a link) so I didn’t see it earlier, so thanks, good tune!

18

Belle Waring 12.16.13 at 1:59 pm

Wait, not identical, it’s “ni yo” vs. “ni no” on re-listen, but still they had to mess with the stress to make it sound like a comforting answer.

19

jackd 12.17.13 at 2:42 pm

If your kids are interested in Vocaloid, they may also want to mess around with UTAU, a freeware alternative. UTAU allows you to record your own voice banks, and the user community has lots of downloads available. The software is in Japanese, and you have to run your Windows machine in Japanese locale.

20

David J. Littleboy 12.17.13 at 3:30 pm

They even have some real people singing over here…

21

Anonymous 12.19.13 at 8:51 am

“Our daughter would want Mandarin/English/Japanese, that could be difficult, and although there’s probably a Japanese bilingual edition I think it’d have all its controls in Japanese and be a pain.”

An English version of Miku’s software came out a few months ago. It’s still expensive and not easy to use, but it is in English now. You can also manipulate things directly in the Edit modes of the “Project Diva” video games, which are still hard to use, but at least don’t cost $150. (And the games are fun, if you like rhythm games at all. “Project Diva F” got an English release for the first time this fall, and is coming out in English for the Vita soon.)

Mikumiku Dance is a freeware animation program that’s commonly used in NDD videos; the quality isn’t quite as good as the ones Sega does (such as the “Rolling Girl” performance you linked), but it can turn out some impressive stuff if you have a lot of time to devote to manipulating 3D puppets. (It’s not Crypton that does the professional-looking 3D animations; it’s Sega. The “live” tours are basically promotional vehicles for their “Project Diva” game line; for the first few tours the link was explicit in the title of the shows. Now they’re more broadly promoting the vocaloid brand, but it’s still Sega that does the technical work for them. You can find traditional-looking 2D animation in Miku PVs pretty easily if you look on NND; the limit on who can do quality cel-based animation isn’t money so much as time, and a lot of NND posters have way too much time on their hands.)

“Vocaloids 02 and 03 are Rin and Len Kagamine, who are blond twins, one girl and one boy respectively, now with 75% less Blood of the Walsungs. OK, 70%. 04 is…some dude. Our daughters have an aversion to the synthesized male voices (which appear to be possibly—no, pretty much we know this, they are sampled, but by artists who don’t wish to reveal their names because then people will be able to ‘force’ them to sing anything they want.) 05 is Luka, she has some good songs, 06 is Gumi, then…”

Some of this is wrong (on the internet, and so I must correct it!1). Rin is vocaloid 02, as indicated by the tattoo on her arm; Luka is 03, as indicated by the same. Meiko is often depicted with a “00” tattoo on her arm, because she came out before Miku did; as far as I can tell her official portraits don’t include it. Kaito (the blue dude with the scarf) and Len don’t have tattoos, because they have long sleeves, and the tattoos are where these numbers come from. But if they did get numbers, it wouldn’t make sense for Kaito to be numbered 05 — he came out before Miku did, around the same time as Meiko. I’ve seen Len marked as 02 and Kaito as 00, both of which make sense to me: they came out alongside the female vocaloids with those numbers.

After those five (which are the five that Project Diva makes use of, and so the ones Sega has made models for), things get complicated: lots of “vocaloids” exist only as fan-made 3D models on NND (like Yowane Haku, whose gimmick is that she sings songs made by people who are bad at using Miku’s vocaloid software), or are third-party programs (like the UTAU-oids, whose gimmick is that their software is free — you can provide your own voicebanks, even), or were made by Crypton but have small fanbases (Gackpoid), etc. There’s a big gap between the main five and the rest, going by NND views, merchandise availability, fanart on Pixiv, or any other metric you choose. And the very idea of “canon” here becomes immediately murky: Miku has a leek because of Ievan Polkka, Rin is associated with steamrollers because of a Jojo’s Bizarre Adventure parody video, Luka is associated with tunas because of a pun on her name… it’s internet silliness the whole way down. Some of it gets picked up by Sega in their games/shows, some of it doesn’t; some of it gets picked up in various manga lines, some of it doesn’t. And the only anime based on this stuff is “Black Rock Shooter”, which is an original property based on the Miku song “Black Rock Shooter”, which was based on a fan-drawing of Miku with one eye on fire and with a weird giant gun — none of the normal vocaloids exist as such in the BRS universe.

Len and Rin both had their voices provided by the same person, Shimoda Asami; in some of their duets it’s really obvious that there’s one person doing both voicebanks. There aren’t any dudes hiding their identities behind male vocaloids: Gackpoid was explicitly based on GACKT’s voice, for example. Kaito and Meiko weren’t based on single voicebanks, which is why there’s no one credited for either of them. Miku, Luka, and the Kagamine twins all have vocals which are produced by manipulating phoneme voicebanks provided by single voice actresses. Kaito and Meiko have a mix of purely synthetic phonemes and uncredited voicebank sources; because they weren’t designed with the intent of becoming pop idols like Miku was, there wasn’t any reason to try to make them seem “real” by linking their voices to a single human source.

“Are CT readers all, like, ‘oh yeah Vocaloids, way to hit us up with some shit from 2010, Belle’?”

2007. I remember when the leeks first became a meme: Ievan Polkka was originally a Bleach flash video, before someone made a Miku version. Aaaaaaand this is the point where I become ashamed enough of how nerdy this comment is that I remove my name from it.

22

heckblazer 12.19.13 at 7:28 pm

I’m reminded of the time the Gorillaz performed live at the Grammys with Madonna. Which was a nifty trick as the Gorillaz are cartoons.

Comments on this entry are closed.