Slashdot Mirror
Researchers Create an Automatic Backup Band for Singers
Researchers at Microsoft Labs are hoping to allow untrained singers to have their own automatic backup band in the near future. A new piece of software, "MySong", promises to take a sung melody and using a probability computation algorithm, generate an appropriate chord accompaniment. There is also a video of the process on the Microsoft Labs website. "'The idea is to let a creative but musically untrained individual get a taste of song writing and music creation,' Morris told New Scientist. 'There was nothing out there that could take a sung vocal melody as an input and then generate appropriate chords to accompany it. [...] Since people rarely sing at precise frequencies, MySong compares a sung melody to the 12 standard musical notes. It then feeds an approximate sequence of notes to the system's chord probability computation algorithm. This algorithm has been trained, through analysis of 300 rock, pop, country and jazz songs, to recognize fragments of melody and chords that work well together, as well as chords that complement each another.'"
Fuck this. This is just going to make pop music even MORE dreadful to those of us who actually appreciate the artistic quality of music. Oh, look, some blond whore can screech into the mic and it'll make the whole damn song for her! Yay! Yes, let's take one of the most important part of music creation and base IT off of a formula now, too.
Show this to your friends and family that don't know what a real hacker is
Now they just need to have artificial voices sing music, and random word generators to make lyrics, and the music companies can stop paying those pesky artists!
Uh, so don't listen to it... I don't understand why people like you get so angry over pop music, welcome to the free world where you can listen to whatever music you like, and hopefully let others do the same.
...even if you can get it to create long, coherent chord progressions, it still will have to stick to chords that match whatever was sung. Even if the system knows how to do jazzy chord changes and secondary function chords and such, an amateur singer won't sing a melody that will flow well with that.
The melody and the chord structure fit together very intimately. If someone doesn't "hear" the chords they want in their head, they probably won't sing a melody that will need an interesting chord progression behind it to make it work.
And of course, for any given melody, there are multiple possible progressions (do you want a IV or a I chord here? Or maybe a V7/V?). The singer will need to have the musical sense to choose which one they want.
All music is fomulaic. How do you think Beethoven was able to compose music even after he went deaf? Bach's music has been used as an example of how the mathmatics of music is similar to Godel's theory. Ever seen two musicians "jam" together? Ever wonder how they can sound so great even though it's the first time they've ever played together? It's becuase it's formulaic, once you know the formula for jazz you can play with any jazz musician (assuming you can play an instrument). Get off your ignorant high horse. Our ears are trained to favor harmonious sounds and reject dischord. There are alot of very talented "pop" artists. The only reason they are considered "pop" artists is becuase their music became popular! That means alot of people like it, it doesn't mean it's crappy music.
Exactly. People seem to have missed the point that this is an interesting innovation in expert systems, not a request to hold forth on how whatever dreary, droning indie crap they listen to makes them superior to everyone around them.
What I'm listening to now on Pandora...
Creating complex music (not 1-4-5 stuff) is actually rather hard. Deceptively so, in fact, because it doesn't always sound hard until you actually try to write it and you realize that nothing sounds the way you thought it would.
As a musician, I actually think it's kind of neat. I could actually have some fun with something like this.
Machines cannot replace musicians. Music is emotional. It is improv. It is creative. Machines do as they are told, and even if they have some complex AI going on, they can still only function according to the parameters they are given. And since a human has to program the machine, the machine cannot be a better musician than the person or persons who programmed it. There is a difference between playing chords to a song and making it your own. Think of all the jazz standards, for example. How many different versions are there of, for example, Misty? Countless. Or how many songs use Gershiwn's "Rhythm" changes? Check this out: http://songtrellis.com/changesPage. Lots of chord changes there. But each version of each song is unique. Music is art. It's not about who is technically "better" or who plays the changes perfectly; oftentimes it the deviations from perfection that can make a song so compelling. Until someone makes a machine with the ability to improvise in response to the lead singer or soloist, convey emotion, *enjoy* music, and discover new things through taking risks and making mistakes, musicians won't become obsolete whatever that means, as if people won't still enjoy making music even if machines *could* do it better.
It's a neat toy, and nothing more. And if crappy pop music uses machines for a backing band, who would even notice? With that form of music, the background music is like the tires on your bike, you don't care about them until they blow. The teenies who buy that crap don't care about music, they are buying into a fantasy that they can be cool and popular and all the crap the pop icon represents. I'd bet that the musicians who back the likes of the Backstreet Boys and Britney and so forth hate it anyhow, they are probably being paid well to be musically bored to death. I feel sorry for those guys. It'd be such a drag to back up a bunch of no talent rich kids. Now, that's a perfect job for machines. Automate the mundane, do the interesting stuff.
blah blah blah
As a music teacher, it is your paid duty to believe and repeat to students that everyone can play music. Have you ever considered the possibility that this is a lie? Have you ever considered that some people are made for certain things and others aren't? Everything is not for everyone. Singing karaoke makes someone a singer as much as Guitar Hero makes someone a guitar player as much as Madden 08 makes someone a football player.
There's this notion of talent and notion of passion. These two go hand in hand. A person has a talent for music, and it is their passion that carries them through to bring this talent to fruition. Does it take a fancy piece of software to help someone discover this talent? Last time I checked, singing started with an open mouth and an "Aaahhh."
Is it elitist for me to believe this? I don't think so. Yes, I am a musician. I started with the clarinet in the 2nd grade and became the youngest player in advanced band. When my friend was trying to learn/play trumpet in 5th grade, he struggled with Mary Had a Little Lamb after several weeks; I picked it up once and played it through. Since then, I learned guitar, bass, piano, and drums. I only took lessons for guitar, the rest are self-taught.
Yes, it takes time and practice, but it also takes talent, and I believe this is something that is natural. I've tried different sports but was never any good. My dad has tried to teach me about cars but I've never had any interest. In an increasingly isolated world, we're starting to forget that everyone in society fits a certain piece of the puzzle that makes up that society. We try to be everything all of the time when we don't need to be. If my car breaks down, I can have my dad work on it. In turn, when he can't fix a computer problem, he can turn to me. When a guitar player needs a beat, he can meet up with a drummer and jam. When a singer needs backing vocals, he or she can meet like-minded individuals and form a group. We have the internet now to make this more possible, yet we seem to be going the opposite direction and disconnecting from everyone else, trying to forge our own miniature frontiers in our homes and apartments.
Let the singers sing, let the musicians play music, let the athletes play sports...we are individual parts of a whole, not whole individuals.
Oh, don't be so quick to dismiss the humble I-IV-V.
Sure, it's easy to pick up and make a song with it. Sure, pop, rock, and just about everything else is built around the I-IV-V. That doesn't dilute the power of the I-IV-V. There's a lot to be said for taking a simple canvas and working with it. It comes from the blues and evolved to jazz (in the form of ii-V-I). It's a formula, and one that you can squeeze a lot out of. I could point you to some songs that use a I-IV-V that would blow you away. You probably could too. The V-I is a natural resolution and that's really what the I-IV-V is built around. I can think of loads of songs that are really just V-I and accomplish the same effect (there's a reason the IV is called the subdominant). In fact, show me a song that *doesn't* use some inversion of a V-I turnaround at least once and I'll show you a song that goes nowhere.
The I-IV-V is like bread: it's been around forever and ain't going anywhere. It's a staple. Don't hate the recipe just because some people can't cook.
blah blah blah
It's a simple function of a limited resource. Only X many albums can be reasonably marketed in a year. The more shit acts there are, the less room that leaves for respectable artists.
I'd much rather have algorithmic vocals than algorithmic music, but I don't sing so I'm clearly biased.
-Billco, Fnarg.com
I am a Music Information Retrieval researcher, and I would emphasize that while programs do not "create from scratch" yet, neither do humans. The notes do not come from the ether, and all that life experience plays a role (I'm also a composer, so I can verify this, too). Our programs do not yet have a lifetime of experience with music, so perhaps it is not a fair comparison?
We don't have all of those things yet, but we do have many parts of this, and we had it a long time ago (check out the works of Robert Rowe and George Lewis for instance). George Lewis' software had a listener and could make musical decisions about what he was playing, as well as what it was playing. We don't have a universal improviser for every style, but not every musician plays every genre either. Still, there's more going on that just the commercial stuff...
I would also propose that Turing is a goal. We're quite a ways off, but there are examples: Phil Winsor has had the music generated by his programs used for TV commercials. There's been quite a bit of progress with machine learning algorithms and self-organizing maps in the field of Music Information Retrieval. We have rulesets for expressive performance (KTH and Generative Theory of Tonal Music) that give reasonably human performances of MIDI data, and we're working on combining these systems with analyses of expert performers.
I have no interest in replacing human performers with computers, but I would love to humanize computer performers.
And create. Don't forget you can use that computer to create music.
"Nine times out of ten, starting a fire is not the best way to solve the problem." - my wife