Customizable Robotic Idol Hatsune Miku

September 18, 2007 8:05 AM

Hatsune Miku is the latest singing sensation to sweep Japan. No, she's not a new idol singer, she's Yamaha's Vocaloid2 software simulating the voice of vocalist Saki Fujita. Currently a #2 seller on Amazon, even at the cost of 15750 yen (about $137). But you don't need to buy the software to appreciate it. Check out Ievan Polka, Fly Me to the Moon, the theme from Princess Mononoke, and more!
posted by CrunchyFrog (40 comments total)

This has big implications. Musicians who aren't singers will be able to create their own vocals in the studio. I can imagine the day when there will vocal style options. You might be able to have a digital Aretha singing your songs, for instance, or whomever. When does the english version come out?
posted by wsg at 8:21 AM on September 18, 2007

Why does she wave an onion around?
posted by Citizen Premier at 8:27 AM on September 18, 2007

Why does she wave an onion around?

Wikipedia to the rescue!
posted by CrunchyFrog at 8:33 AM on September 18, 2007

Holy crap, it's Sharon Apple.
posted by gurple at 8:36 AM on September 18, 2007

I wonder if this is what it sounds like when you stand at the edge of the uncanny valley and yell, "reverb!"
posted by Robson at 8:36 AM on September 18, 2007 [3 favorites]

Gosh, what a catchy tune, that polka.
posted by Wolfdog at 8:40 AM on September 18, 2007

The original Vocaloid got an English version, so I imagine that its top-selling followup will be translated as well. No idea when, though.
posted by CrunchyFrog at 8:43 AM on September 18, 2007

You know, every now and then I have a moment where I find myself living in a cyberpunk novel. A friend of mine brought this up recently and ever since I've been spotting the crossovers more and more.

This is one of them. Rei Toei by any other name, and all that.

No wonder the latest William Gibson novel is set in the present. Why the hell bother making up a future when you get maybe ten years out of it before it becomes outdated?
posted by Jilder at 8:46 AM on September 18, 2007 [2 favorites]

It's worse than that. Reality passed us a while back. That's why we're all reading steampunk.

At this rate, the next big meme will be flintpunk.
posted by sebastienbailard at 8:50 AM on September 18, 2007 [1 favorite]

Wow, I didn't know we could do this. Good find.
posted by hjo3 at 8:52 AM on September 18, 2007

You know, every now and then I have a moment where I find myself living in a cyberpunk novel.

Hell, forget Gibson. Norman Spinrad wrote about this in 1987.
posted by JaredSeth at 8:59 AM on September 18, 2007

That version of Fly Me to The Moon is truly awful. While I enjoy much of Japanese culture and art, their love of saccharin cute in any form is hard for me to fathom.
posted by doctor_negative at 9:13 AM on September 18, 2007

doctor_negative:

In Tennessee, a guy sees a sign in front of a house: "Talking Dog for Sale."

He rings the bell and the owner tells him the dog is in the backyard. The guy goes into the backyard and sees a black mutt just sitting there.

"You talk?" he asks.

"Yep," the mutt replies.

"So, what's your story?"

The mutt looks up and says, "Well, I discovered this gift pretty young and I wanted to help the government, so I told the CIA about my gift, and in no time they had me jetting from country to country, sitting in rooms with spies and world leaders, because no one figured a dog would be eavesdropping. I was one of their most valuable spies eight years running."

"The jetting around really tired me out, and I knew I wasn't getting any younger and I wanted to settle down. So I signed up for a job at the airport to do some undercover security work, mostly wandering near suspicious characters and listening in. I uncovered some incredible dealings there and was awarded a batch of medals. Had a wife, a mess of puppies, and now I'm just retired."

The guy is amazed. He goes back in and asks the owner what he wants for the dog.

The owner says, "Ten dollars."

The guy says, "This dog is amazing. Why on earth are you selling him so cheap?"

The owner replies, "He's such a liar. He didn't do any of that shit."
posted by small_ruminant at 9:30 AM on September 18, 2007 [19 favorites]

Musicians who aren't singers will be able to create their own vocals in the studio.

Thanks to ProTools, that's been happening for years.
posted by miss lynnster at 9:33 AM on September 18, 2007

I learned to speak Japanese from this vocaloid program and couldn't figure out why people were laughing at me.
posted by fleetmouse at 9:55 AM on September 18, 2007 [3 favorites]

A friend of mine brought this up recently and ever since I've been spotting the crossovers more and more.

Your friends live journal essay is pretty annoying.

<sub>Not so long ago humanity hit an incredible milestone: we landed a robot on Mars and it took video of the planet's surface from a first-person perspective. We had R2D2's happy snaps of an alien world! It looked like the Australian outback, but wrong. All purple skies and utter vacancy. Personally I was entralled just by the rocks, rocks which had lain there for thousands of years, while down on Earth we were busy having history.

But no, that's not the focus here. The focus is that when the Mars landing happened, it was relegated to a sidebar on page three of most of the newspapers, while Paris Hilton's feud with Nicole Richie was headline material.</sub>

It wasn't big news because it wasn't the first time we've done that. The only difference between this and the stuff done in the 60s and 70s was that it could wheel around.

He also calls an 8-year old talking on a cellphone 'dystopian'. Okay...
posted by delmoi at 10:31 AM on September 18, 2007

Musicians who aren't singers will be able to create their own vocals in the studio.

They could spend hours and days of their valuable time producing something that approximates the subtleties of real vocal tonality and phrasing, those elusive elements that get called "soul". Yes, indeed they could. Or they could just, you know, hire a session singer.

The "digital Aretha" is maybe closer to the mark in the long run, but how many years is it going to be before such software is producing something equivalent to a decent vocal impersonator, without having huge amounts of time invested? In terms of phrasing in an original song, for example?

Up to this point in time, it has always been easier to get a good take of the real thing in the first place than fuss around in the studio later because someone can't do their job properly. I don't see that changing soon.
posted by howfar at 11:04 AM on September 18, 2007

I cannot wait for her come back album.
posted by GoodAaron at 11:28 AM on September 18, 2007

This is kind of cool, but why are there human-like limitations on the software? From the excerpts I listened to, the software stays within the range of a normal female voice, for instance. As howfar asks, why not just hire a session singer for that? Once the novelty wears off, a real singer (who's good) is always much better.

I want to hear software that's like a human voice, but does things humans can't do, like sing 5 octaves. It doesn't make sense to create this stuff and then impose human limitations on it.
posted by LooseFilter at 11:30 AM on September 18, 2007

This use of technology is so mundane. It's like using a sampler to make the barking dogs sing jingle bells. Someone needs to do something fun with it, like install it in Stephen Hawking's wheelchair. "And now we talk about first second of universe desu!"
posted by fleetmouse at 11:33 AM on September 18, 2007 [3 favorites]

LooseFilter, but what would the real innovations be? You can artificially give a human singer 5 octaves already, but it's not really all that exciting. Timbre and timing, those are the really interesting textural elements in pop music. I'm excited by any application of technology that that does innovative and satisfying things with them. But that could be any technology, no matter how old.
posted by howfar at 12:03 PM on September 18, 2007

Or they could just, you know, hire a session singer.

They don't even need to do that. You can be plenty good with bad vocals.

To paraphrase Tolstoy, good vocals are all alike. But bad vocals - man, they can be bad in entirely unique ways. It's not all Leontyne Price and Marvin Gaye. People still listen to Aesop Rock, or Bob Dylan, or Bright Eyes, or Slayer, or Louis Armstrong, or Motorhead, or, hell, Wesley Willis. Not pretty. Distinct.
posted by kid ichorous at 12:49 PM on September 18, 2007

I'm quite fascinated by this, actually. The voice doesn't sound quite right, and yet for me, it has a certain charm--not the saccharine-sweetness referred to doctor_charm, but rather the pseudo-electrified and faintly distorted voice. It's like a faint accent.

Anyway, some more information on the "..and more!" links, if anyone's interested:

"...and..." is a version of "Hare Hare Yukai", originally sung by HIRANO Aya, CHIHARA Minori, and GOTO Yuko, as the ending song in the anime "The Melancholy of SUZUMIYA Haruhi". A re-enactment of the dance that accompanies this song actually exploded across Japan, the YouTubes, and anime conventions, in a manner not unlike what happened to Gary Brolsma (who followed along to O-Zone's "Dragostea Din Tei").

"...more!" is a version of "Uninstall", originally sung by ISHIKAWA Chiaki for the opening of the anime Bokurano, which debuted spring of this year.

(Lots of YouTube/Wikipedia.)
posted by qcubed at 1:00 PM on September 18, 2007

People still listen to Aesop Rock, or Bob Dylan, or Bright Eyes, or Slayer, or Louis Armstrong, or Motorhead, or, hell, Wesley Willis.

Well yeah. But all their records had good vocals. No-one here seems to have made any statement about the kind of vocal (one's own or anyone's) one might wish to either hire or synthesis. I while I agree that many non-pretty vocals are good, no-one should be held to be morally obliged to sing if their voice isn't the instrument they want to use.
posted by howfar at 1:10 PM on September 18, 2007

*sigh* I can sing just fine. I can write lyrics, too. But I can't compose music. When will someone create some software that will let me select a few options and then spit out an original tune?
posted by Faint of Butt at 1:29 PM on September 18, 2007

Oh they have that now too. It's called Sequel.
posted by fleetmouse at 2:28 PM on September 18, 2007

I really don't think that human singers have any stiff competition for a long time yet. The voice is way too nasal, and has terrible phrasing.

That said, I'm still loving the Loituma - now there is some great singing, from all members of the band, but especially the scat artist.
posted by jb at 2:29 PM on September 18, 2007

Technically this is really neat, culturally rather interesting, musically not that great, at least for now. Here's a demo of another Vocaloid 2 library, Sweet Ann, against a real female singer. Extremely high notes do not sound like singing at all.
posted by ikalliom at 2:51 PM on September 18, 2007

I don't specifically disagree that a human singer is a better choice all around, but a lot of the arguments I'm seeing here really remind me of the kinds of discussions that came up when it became trivial to bypass a living drummer and go with an electronic solution.

The technology isn't there yet, but in ten years who knows what they will be doing with this.
posted by quin at 2:57 PM on September 18, 2007

Oh, my. I was ready to scoff at Sequel, on the grounds that I don't need to compose any Rock or "Electro," but I think that might actually be able to spit out a waltz in D minor if I told it to. Perhaps more research is in order.
posted by Faint of Butt at 3:01 PM on September 18, 2007

The challenge is that the human voice has an "uncanny valley" too. Those last five meters of incredibly deep gap will take a long time to bridge, because you'd basically need a culturally literate AI to pull off emotional and contextual nuances without hundreds of hours of hand-tweaking.

It's this way with orchestral samples too. You can record all the articulations you want with as many transitions from note to note and playing style to playing style as you can think of, but a dedicated instrumentalist knows WHY, not just how and when, to use those techniques.

There is a program called Synful Orchestra that does a similar thing with solo instruments - violins, flutes, etc - it's pretty amazing to hear it use vibrato, legato, marcato playing automatically when you just play the keyboard, but it's in the same "neat trick, but not convincing" boat.

I'm terrified because I'm sure some insane Yamaha genius already has it figured out and is planning to install it in death robots. They will sing lullabies (optional Wagner setting) to you as they approach, arm-scythes at the ready.

Anyway, hooray for music technology!
posted by jake at 3:11 PM on September 18, 2007 [2 favorites]

Here's a much less annoying demo of Sweet Ann. I'm more partial to Lola myself (though Leon isn't without his charm).

The technology isn't to the point where I'd want to release anything using it, but it would be great for someone like me to show a vocalist exactly what I want, particularly when collaborating online.
posted by malocchio at 3:37 PM on September 18, 2007

quin, I think your point is really the key to this discussion. Electronic drum programming has made things possible that weren't before. People have done brilliant things with it, and created whole new genres of music because of it. It's not really a terribly good choice, however, if you are looking to replicate the particular strength of a human percussionist. Same with this. It may be used in marvellously creative ways, and those applications are the musically interesting thing, not the possibility of disposing of human singers. In the same way, drum programming hasn't eliminated human drummers. Although who hasn't wanted to eliminate a few human singers from time to time?

This is an extremely sophisticated synthesiser. The latest in a long line of synths stretching back for over a hundred years now. Synths have certainly had real effects on musical history, and we would be poorer without them. In the end, however, it is what people do with such tools that matters, not how cool the tools are in themselves. I don't mean to disparage tools or toolmakers (Stradivari made tools!) but to point out that tools are, ultimately, valuable because of the end to which they are put.
posted by howfar at 3:45 PM on September 18, 2007

fleetmouse: That would be worth it, if only to see Stephen Hawking in a gothic lolita dress.
posted by Saydur at 4:18 PM on September 18, 2007

LooseFilter, but what would the real innovations be?

I don't know--that was just the first thing that sprang to mind (had to run off to rehearsal, actually).

Electronic drum programming has made things possible that weren't before. [...] In the end, however, it is what people do with such tools that matters, not how cool the tools are in themselves.

Yes, that's what I was trying to say. So, repeated for emphasis. Should this software catch on, it will likely follow the path of most musical technological innovations--someone(s) create the tool, and other people come along and figure out amazing, unexpected things to do with that tool. I think of people like Stoelzel, adding early valves to brass instruments, and others who refined, improved, reworked; they didn't think of the fantastic things to do with fully chromatic brass instruments, the composers did.
posted by LooseFilter at 5:35 PM on September 18, 2007

So that's why this girl is dancing with an onion. Warning - sexy.

thanks crunchyfrog, this is cool.
posted by vronsky at 5:37 PM on September 18, 2007

woops
posted by vronsky at 5:38 PM on September 18, 2007

Oh, Japan! What won't you dehumanize?
posted by stavrosthewonderchicken at 6:28 PM on September 18, 2007

Dear Britney Spears,

See what you've done! Surely, Autotune was a clever way to inch past the stringent talent barriers put up by the music industry, but look at the slippery slope down which it's led.

Those folks that reanimated Gene Kelly (MOV) and Orville Redenbacher will soon be coming for your voice!

Don't say I didn't warn you. Best to the boys.

xo,
-Vulcan Mike.

p.s. This is why everyone's all up in arms about IP laws and copyright now. It would be nice to get the public's interest squared away before the first case of an artist signing their voice over to a label, receiving pennies and watching as it's used for billions of dollars worth of artificial music.
posted by VulcanMike at 9:16 PM on September 18, 2007

Vulcan--

Britney's voice was never the product.
posted by effugas at 10:27 PM on September 18, 2007