Slashdot Mirror
Synthesized Singers
ctwxman writes "Over the past few decades, advances in computer hardware and software have eliminated many jobs... some technical, some menial, but none artistic. As an on-camera performer in television, I've always was believed that I was 'bulletproof' as far as replacement through technology was concerned. Not so fast. Recently, The Sinclair television stations began using 'central casting' to bring news and weather anchors from a central location (near Baltimore) to the local outlets. Still, real people are needed, just not as many. But now, even real performers may be replaced. The New York Times (inhalation of airplane glue required) reports on a new technology which allows synthesized singers to sing. Imagine having a singer with a world-class voice at your disposal, any hour of any day. She's just standing at the ready, game to perform whatever silly song you might make up for her: a ballad about her love for you, a tribute to your best friend's golf game, a stirring rendition of the evening's dinner menu. Scary."
What is really cool is when machines can do what people cannot do. The first sign of this was several decades ago when drum machines and analog synthesizers came about. The drum machines could play beats so fast and hit more instruments simulateously than a single person has limbs for and the synthesizers could create entirely new sounds. In the present, there are pitch machines which put singers' voices at a desirable pitch when singing. Hopefully next we'll have robots/machines with AI that can create their own insightful, fun, or intelligent lyrics to songs and sing them to an original beat. Popular music analyzers(just posted on /.) are already capable of predicting what tunes have potential. Music is a product of man, whether it is created through human hands or machines. You can't mentally hold yourself back to the idea real music is only a direct product of man.
(\_/)
(O.o) This is Bunny. (> <)
Serious legal issues arise when creating "voice fonts" made from singing material previously released by artists. I doubt the RIAA or the artists themselves will like this new techonlogy at all. If this technology is a success then I forsee a push by the RIAA/artists themselves to get their voices copyrighted.
As an example, Harley Davidson (the motorcycle company), tried to get it's unique motorcycle engine sound copyrighted and failed. Will this change the copyright office's stance?
The problem with computer-synthesized voice is that it will not correctly convey emotions, and (if plaintext) will not even stress the right words. And it would be a bitch to write the appropriate tags.
eg
I didn't say he did it.
I didn't say he did it.
I didn't say he did it.
Don't waste your vote! Vote for whoever you want, unless you live in a swing state it won't matter anyways
I write music and produce for TV series. I have never had to use a musician. Ever. My boss uses live performers occasionally for shows that might win Emmys. I use Machfive and Digital Performer 4.1. Samplers, especially the 300 dollar Machfive platform/plugin have eliminated the need for live artists in my business. Hell, I will be recording a rap (bleh) artist soon, and the only live recording will be his vocals. The rest will be sampled.
Your time is coming to an end, but I will say that synths and samplers don't match live studio musicians...yet. Vocalists are still safe, at least until Apple fixes their Speech voices.
Sig (appended to the end of comments you post, 120 chars)
Combine these synthetic vocals with some randomized instrumentals and pipe it into your 'hitablity` algorith (covered here and here) and generate endless pop music!
I want a soundtrack for my life. Like when something goes good, there would be a choir of "hallelujah". So far I only have this site for when I mess up.
it's interesting how the technologies of voice replication and voice recognition are so similar. The recording studio where I work recently participated in a project for the developers of a voice controlled navigation system (think OnStar). Our task was to record people born and raised in Chicago dictating a long list of words which would presumably use at least most of the phonemes at our disposal. I think they did this in most major metropolises. The goal was to build a system that could recognize english speech patterns with a wide colloquial variance. Perhaps it won't be long before we have a program that can emulate anyone's voice.
Did you see the movie "Farinelli?"
m
http://www.medieval.org/emfaq/misc/farinelli.ht
We've been doing a lot with singing voice for a long time now.
I actually develop and work with voice synthesizers that are phoneme and phrase based. We dig up an actor, or an unwitting researcher from another project, and make them rattle off a few thousand sample sentences. Then, in theory, we use that date to make a "realistic" human voice. Yeah. Sure. Riiiiiight. If it worked, I'd be out of a job. :) Technology has come a very long way, but it is no where near what a real person can do. It can't even imitate. The spoken word is a thousand times easier than singing, I'll believe it when I hear it sing what I say within an hour of placing a request.
By the way, I proposed that we develop a rapping version of our synthesized voice, just for fun. The idea was trashed in 30 seconds, b/c who wants a voice that raps like an off duty LAPD Sergeant at Karaoke night? I doubt this thing can sing that well...
"Life's funny sometimes." "And sometimes it isn't." --Cat's Cradle
Classic singers stay as close as possible to the "absolute" quality line - it's perfect for being mathematically modeled and it's a matter of time such models will be apparead, even if their implementations will take some hardware resources.
Pop singers make sound anyway far away from being called as an art. It's perfect for being implemented in embeded solutions. It's a matter of time first cyber-singers will be cloned like cheap "made-in-China" electronic (sorry, my oriental friends, although nothing personal or racial in this comment).
Jazz is still an art, like classical music, but its improvizations are very unpredictable. Jazz singers will be last ones to go. Even more - Jazz improvizators will be eventually involved to prototype new cyber-singers. Hmm, I can even imagine special programming languages for singer-modelling: "bebop", "blues", "swing" :)
Less is more !
Many people here have already commented that the voice isn't quite there yet in terms of realism. Many here have pointed out that the technology doesn't seem new -- they're just taking more sound samples and blending them together (albeit in the frequency domain, with smoothing).
One area that really needs quite a bit more work is the vowels. When singers sing "ee" (as in "saved a wretch like me," for example), they usually soften it so it sounds a bit more like "meh." When I used synthesized voices before on a Mac, I had to specifically spell the lyrics as "meh" so that the program would articulate those vowels properly.
Of course, maybe the sample of "Amazing Grace" I heard was recorded by a singer who really liked to pronounce such vowels as "ee" rather than soften them to "meh," but it doesn't seem likely.
Gan Family Homepage
What about sign painters? There are still 2d animators around, they just migrated media from paper->cel to paper->digital. Of course, those that are still working (no thanks to brainless execs who think 3d is the panacea for bad, overworked and overbudget stories...)
It might actually be easier to do singing than normal speech
Given that singing and speech aren't mediated by the same parts of the brain. In this book, there's examples of people who can sing sentences, but can't speak them.
High on a rocky promontory sat an Electric Monk on a bored horse. From under its rough woven cowl the Monk gazed unblinkingly down into another valley, with which it was having a problem.
The day was hot, the sun stood in an empty hazy sky and beat down upon the gray rocks and the scrubby, parched grass. Nothing moved, not even the Monk. The horse's tail moved a little, swishing slightly to try and move a little air, but that was all. Otherwise, nothing moved.
The Electric Monk was a labour-saving device, like a dishwasher or a video recorder. Dishwashers washed tedious dishes for you, thus saving you the bother of washing them yourself, video recorders watched tedious television for you, thus saving you the bother of looking at it yourself; Electric Monks believed things for you, thus saving you what was becoming an increasingly onerous task, that of believing all the things the world expected you to believe.
Unfortunately this Electric Monk had developed a fault, and had started to believe all kinds of things, more or less at random. It was even beginning to believe things they'd have difficulty believing in Salt Lake City. It had never heard of Salt Lake City, of course. Nor had it ever heard of a quingigillion, which was roughly the number of miles between this valley and the Great Salt Lake of Utah.
The problem with the valley was this. The Monk currently believed that the valley and everything in the valley and arround it, including the Monk itself and the Monk's horse, was a uniform shade of pale pink. This made for a certain difficulty in distinguishing any one thing from any other thing, and therefore made doing anything or going anywhere impossible, or at least difficult and dangereus. Hence the immobility of the Monk and the boredom of the horse, which had had to put up with a lot of silly things in its time but was secretly of the opinion that this was one of the silliest.
How long did the Monk believe these things?
Well, as far as the Monk was concerned, forever. The faith which moves mountains, or at least believes them against all the available evidence to be pink, was a solid and abiding faith, a great rock against which the world could hurl whatever it would, yet it would not be shaken. In practice, the horse knew, twenty-four hours was usually about its lot.
So what of this horse, then, that actually held opinions, and was sceptical about things? Unusual behaviour for a horse, wasn't it? An unusual horse perhaps?
No. Although it was certainly a handsome and well-built example of its species, it was none the less a perfectly ordinary horse, such as convergent evolution has produced in many of the places that life is to be found. They have always understood a great deal more than they let on. It is difficult to be sat on all day, every day, by some other creature, without forming an opinion on them.
On the other hand, it is perfectly possible to sit all day, every day, on top of another creature and not have the slightest thought about them whatsoever.
When the early models of these Monks were built, it was felt to be important that they be instantly recognisable as artificial objects. There must be no danger of their looking at all like real people. You wouldn't want your video recorder lounging around on the sofa all day while it was watching TV. You wouldn't want it picking its nose, drinking beer and sending out for pizzas.
So the Monks were built with an eye for originality of design and also for practical horse-riding ability. This was important. People, and indeed things, looked more sincere on a horse. So two legs were held to be both more suitable and cheaper than the more normal primes of seventeen, nineteen or twenty-three; the skin the Monks were given was pinkish-looking instead of purple, soft and smooth instead of crenellated. They were also restricted to just one mouth and nose, but were given instead an additional eye, making for a grand total of two. A strange looking creature indeed. But truly excell
Wouldn't it be smarter to make a virtual mouth and larynx and make it sing? So that the computer would calculate the resulting voice.
Maybe someone has already tried. Sorry about my bad English.