Slashdot Mirror
Resurrecting Performers Via Computer Performance
putko writes "The NYT has an article entitled 'Play It Again, Vladimir (via Computer)' that discusses efforts to transform old recordings into new, computer played performances (reg. required), by determining how the previous performer made the sounds and redoing it. Further efforts attempt to distill the 'style' of a performer and play other scores with the same style. As can be expected, musicologists argue over whether or not the new musical artifact is really 'a performance'. Philip K. Dick would be proud."
from about a month and a half ago we discussed it longly at my job(music software store)
Muzik.4.Machines
Ask a piano player if a digital piano is a passable substitute. Yes it's pretty damn good... but still not the same...
Sinatra singing Oops I did it Again?
There are some innovations which are novel, but aren't quite built to be of use.
My Linux - (L)ove (I)s (N)ever (U)tterly eXPensive
Yeah but can it do hardcore gangster rap?
__Funny Adult Videos
This displays the power of modern computing. To be able to "replicate" a song by replicating human thought processes shows that, finally, there is a balance between fast systems, and complex software available to utilise them. After all, what use is a 10Ghz 512-bit 3Ghz FSB 1GB video RAM 10GB RAM machine - when you're running Word? Complex simulation programs are the way of the future.
See more information about it here, from (*ahem*) an older Slashdot article...
--- At my sig, unleash hell.
Cutting Archives does a lot of restoration work. Check their faq
We also had a cool story on slashdot before about Concert to be Performed from Beyond the Grave
The humanity is in the variability (think of them as "mistakes" in a flawless performance) not in the perfection. And a perfect imitation of a specific instance of humanity is still not human, because the mistakes are out of context. This is the same reason that random computer-generated mistakes, even perfectly random ones, still don't sound human.
When a human performs, the performance is subtly affected by the things that affect humans: the weather outside and whether it's gloomy or not; the fact that it's the holiday season; the fact that a leader has been assassinated or the performer's daughter has been ill; the musty mugginess of the air in the auditorium... these subtle types of phenomenological data affect human performances in ways that the audience and performer can share as a kind of unconscious communication, at least so long as they are from the same culture.
A computer that reproduces a previous performance, even if it does so perfectly, does so out of context. It is making all the wrong mistakes for the current situation, so it's playing just doesn't ring true. Until computers can feel gloomy because of gloomy weather, or can be thrilled because the millenium dawns at midnight, five minutes from now, they won't be able to produce performances that truly move us in the same way that human performances do, because that element of unconscious situational communication and solidarity in shared experience is missing.
STOP . AMERICA . NOW
Will we see covers of new material by long-departed artists? E.g. Lemon Tree performed by the Beatles.
The music companies would have a field day with this. Push a button and you can have a cover of everything by everyone. Not to sell, just to flood P2p networks.
Hal Spacejock: Science Fiction with Nuts
This is a cute idea. However, without constant communication and feedback between the audience and performer, as well as introspection on the part of the performer, I'd hesitate before calling it a performance, but rather a presentation.
-- I prefer the term "karma escort."
I am a musician, and I have heard these things in person. For performances recorded directly on a Disklavier, the recording is indistinguishable from the original to my ears (and to every other musiciain I have talked to about this). If the technique in the article is indeed accurate, then this could mean great things. However, as the article mentioned, it is much more difficult to determine when the notes stop sounding, and pedalling, than the attacks. There is the interesting question of copyright: for ancient recordings ressurrected, who owns what? and is it possible to just tweak a few notes and then do what you want with the thing? (remember, the piece, and the recording are P.D.)
The toad can't burp - and for some reason can't fart either, so it swells up and eventually explodes. --Anonymous Coward
Yamaha has done this with their Disklavier player pianos, so that you can listen to an artists song as the artist actually played it.
It is neat to look at a nice grand piano playing, without anyone sitting at it, keys moving and everything, knowing that if Gershwin were here to play it himself, it would sound just the same.
That, personally, had far more of an impact than just hearing the same piano play the same song.
Lose Weight and Feel Great with Isagenix
Think of the possibilities! When actors and singers can be recreated through computer reproductions, we could end up putting live actors out of work, or force them to sit in front of a computer or microphone more than they already do to perform in a high-tech production.
We could see a resurgence in the popularity of live stage productions, as people grow weary of computer generated reproductions. And we could see a whole new way of recreating old TV shows like Star Trek. Instead of http://www.newvoyages.com/ having live actors playing the parts of Kirk and Spock, we'd see a computer generated Bill Shatner, and L. Nimoy as they were in the 1960s.
Saskboy's blog is good. 9 out of 10 dentists agree.
At last they recognize superiority of MIDI format.
Indeed, MIDI mixed for surround-sound would be divine musical perfection.
I suggest you read Slashdot
I think Sousa said it best in his comments that recording would kill off live music.
This technology brings up two points:
1: Are we now trying to eliminate the performer from the loop altogether?
2: Have listeners become so detached from good performances, as predicted by Sousa, that we can't tell the difference between a live performance and a replication in the style of one particular live performance from long ago?
-- I prefer the term "karma escort."
Here's some inventions of Philip K. Dick (via technovelgy.com).
The technology described by Philip is definitely not in this list; the article's submitter is either lazy or cleverly attempting to sneak the dupe past the editors via his absurd PKD reference.
Andy - an artificial human
Autofac (Nanorobots) - a factory that can replicate itself
Bubblehead - big-head brainiacs
Claws (Guard Robot)
Commuter Cooling Unit - portable air conditioning
Dr. Smile - psychiatrist in a suitcase
Electric Sheep - livestock as consumer electronics
Embryonic Robots - early scifi nanobots?
Empathy Box - TV for your emotional brain
Extra-Factual Memory - an implanted memory
Homeopape - news just you can use
Kipple - non-recycled paper.
Mood Organ - play your partner
Nanny - child-care robot with punch
Nexus-6 Brain Unit - meet my friend Roy
Penfield Wave Transmitter - an emotional brain remote control
Perky Pat Microworld - playset for grownups
Precrime Analytical Wing - precogs babble, machines tabulate
Replicant - an artificial human
Robot Cab Driver- everybody's got problems
http://www.thebricktestament.com/the_law/when_to_
But if it attempts to produce the famous work of john Cage and his 4:33 - I am sure it will cause the machine to go into a run condition and crash....
Where the mind is without fear and the head is held high; Where knowledge is free; Where the world has not been broken up into fragments by narrow domestic walls; Where words come out from the depth of truth; Where tireless striving stretches its arms towards perfection;
Rabindranath Tagore
Where The Mind is Without Fear
Gitanjali
Creed is broken up. Let's keep it that way. There will be enough terrible bands in the future. The last thing we need is future generations resurrecting our terrible bands.
Why, all we have to do is lobby to extend copyright for *another* fifty years, and we can keep milking more money out of long-dead artists, now using their acts even after their body has given out!
Any program relying on (nontrivial) preemptive multithreading will be buggy.
This is a "digitally resurrected recreation" of a previous slashdot post.
Clearly your post should have been entitled "Play it again Slashdot"...
I've heard one of the Telarc recordings of Rachmaninoff reconstructed from player piano rolls. It didn't do a whole lot for me, but then again, it might be that I don't really like how Telarc discs (or Deutsche Grammophon, for that matter) are engineered...
This sig intentionally left justified.
re-create dead actors in the flesh...check...(SW movies)
re-create dead performers playing...check...
Hmm...what exactly do we need people for? W00t!!! Death to the humans, computers conquer all...wait a sec *looks down*...hey...I'm real flesh and blood...uhoh
All the Dicks in the RIAA, who won't have to pay those pesky royalties. Maybe they can finally elevate their standard of living to that of music producers...
Oh, wait...
Faith: n. -- That human impulse that drives them to steal appliances when the power goes out
When there's an article on NYT, and you want to read it without registration, just search for the title in Google News. Once you have Google News as the referrer, you get free access to the first page of the article - which is usually enough.
I just read the summary, not the articl, but it immediately reminds me of CPU Bach on 3DO. I wasn't a big Bach fan, and still prefer others, but it was pretty fascinating to see how it would compose classical music while explaining what it was doing and how the different parts were supposed to work.
Interesting, yes. Fascinating, yes.
But... a concert pianist would not play the same work twice the exact same way. Interpretive training from a decent conservatory doesn't turn out robots. Room dynamics (acoustic, audience, time of day, etc.) often contribute heavily to a perceived "good" performance.
In fact, such a synthesis program should really include an improvisatory component -- a "learning" program that offers slight deviations in appropriate moments. Of course, learning what those are is probably less quantitative than such a system might allow...
Account #1
login: wisterian
password: wister
HD Trailers
Here is a pretty cool article on a similar project, but from the software development point of view.
--- Hot Shot City is particularly good.
I guess if it works, it could reproduce the music exactly, since the music was probably done on synthesizers equipment anyway.
....
Some sounds have not yet been synthesized very well: take the nuances of bowed violin sound for example (a succession of transients). Still a long way, I'd guess, from re-doing an early performance of Joachim or Busch, or even a more recent but still early Heifetz or Menuhin performance
-wb-
I'd like to see how the software the analyzes the music would compare to a piano roll or Disklavier recording. Have a musician play a selection on a Disklavier and record it. At the same time, record the audio and use the s/w to try to replicated it. Play both the MIDI recorded version back as well as the reverse engineered MIDI version and compare both to the original recording.
I bought the "Gershwin Plays Gershwin" http://www.keyboardwizards.com/billboar.htm CD a few years back and really enjoyed it. Copious liner notes told how they restored the rolls, fed them into a 1911 Pianola that "played" the Disklavier which was recored as a MIDI file. Other rolls were scanned, then converted to MIDI files. You can buy these MIDI files and play them back on your own Disklavier if you want.
One track on the CD was a contemporary audio recording, letting you see that for the time, piano rolls really were the way to go for quality. Like MP3s today or cassettes a while back, cheap and easily distrubuted beat out quality for wide distribution.
The most surprising thing in the notes was the history of creating the piano rolls. Initially they were created on special pianos that cut the rolls as the selection was played. These were played by an actual piano player. But rolls were also created by piano roll cutters who basically took an exacto knife to the roll. Besides being cheaper to employ, these cutters also didn't accidently hit a wrong note in the middle of a peformance. It was also possible to create rolls that could not be played by a normal human with 10 fingers ( Shades of Gattaca here ). So in the early 1900s technicians were doing essentially the same thing as someone typing out MIDI code and then feeding it into a Disklavier today.
Do we really need computers taking over the arts too?
I'm wracking my brains for lasting excuses to value human performances over computer performances. You know, the kind of excuses that keep working no matter how good the computers get. AI is supremely cool stuff, but when I imagine the issues we'll inevitably have to deal with -- this century or next -- it scares the hell out of me. I'm gonna go struggle with my moral dilemma now.
Mrs Schmegma, Stand up Please
OSGGFG - Open Source Gamers Guide to Free Games
You might be interested in this article from the New Yorker, about the effect recording has had on the way classical music is performed. The gist of the article is that before classical musicians could hear recording of themselves and of interpretations from other countries, there was a lot more regional variability, and a lot less emphasis on perfection. Now that studio performances can be digitally tweaked or spliced together, the emphasis on perfection is even greater.
Edward Elgar's recordings of his Second Symphony and Cello Concerto, from 1927 and 1928, respectively, are practically explosive in impact, destroying all stereotypes of the composer as a staid Victorian gentleman. No modern orchestra would dare to play as the Londoners played for Elgar: phrases precipitously step over one another, tempos constantly change underfoot, rough attacks punch the clean surface. The biographical evidence suggests that this borderline-chaotic style of performance was exactly what Elgar wanted. "All sorts of things which other conductors carefully foster, he seems to leave to take their chance," a critic observed. Modern recordings of Elgar are so different in sound and spirit that they seem to document a different kind of music altogether.
to make old recordings new...Glenn Gould without the mumbling
Why? Part of what made Gould's performances so special was the fact that he did mumble during them. Hearing him mumble helped you understand his mind. I say it isn't Gould if I can hear any mumbling!
This sort of reminds me of the work of David Cp[e, who's EMI (Experiments in Musical Intelligence) program uses a sort of intelligent recombination to create new compositions based on a composer's existing works.
The scheme has three basic components:
- Deconstuction (analyze and separate into parts)
- Signatures (commonality - retain that which signifies style)
- Compatibility (recombinancy - recombine into new works)
. EMI is constantly evolving, so the source code isn't available, but a simplifed (but still quite effective) version of EMI called SARA (Simple Analytic Recombinance Algorithm) is has been released which creates new compositions by replacing each measure in an existing work with a functionally equivalent measure (but different) measure found in the same composer's works. It's pretty darned effective.Cope has been able to create convincing reconstructions of Bach, Beethoven, and even Gershwin.
One video I've seen showed EMI with a 'Beatles' option in the menu, but he's never released any Beatles examples, so I suspect that 'sound' of a pop group relies more on features such as the performer's voices and instrumentation than a particular 'style' embedded in the music itself.
Interestingly, while EMI generates MIDI output, it's much more convincing to have a human play it back. It would be interesting to hear computer-generated 'Gershwin' player interpret Cope's computer-regenerated 'Gershwin' compositions.
If he were alive today he might very well choose to play his performances via a a synclavier.
As a musicologist (technically music theorist but eh), I would view the new reconstructions as performances. To be really clear, I should say that I would view the process of reconstruction and the event as a whole as a performance. If one considers the "original" music (the first established instance of a score for example) the work, then one performance of the work is an interpretation and a recontextualization of the interpretation is a meta-performance. If we all disabuse ourselves of the notion that there is such a thing as "the work" as an absolute, however, then these computer reconstructions are direct performances on the material they have chosen, which just happens to be a performance itself. No, it won't be "authentic" and it won't be "the same" and even if it were, that wouldn't be the point. The performance is not in the end result but in the process of recontextualization.
Now don't make me whip out literary theory.
"The Cube": it just wouldn't be the same without fellatio "Corey Kosak": It just wouldn't be the same... oh, looks like
A digital piano consists of the sounds of a piano incorporated into a silicon chip. Either as a recording or sampled recording of the actual instrument, or the parameters needed to drive a synthesizer into producing a piano sound. Every year brings the chips closer in sound to the real instrument.
What makes the chip a 'passable substitue' for a piano is that a chip weighs at most a hundred grams while a piano weighs several hundred pounds.
And, a chip that reproduces the sound of a piano costs at most a few dollars to manufacture, while a real piano costs many hundreds of dollars to buy.
Only rich musicians with well-muscled roadies would claim that the-piano-on-a-chip is not a passable substitute to the real instrument. The technology for a real piano is fixed, but the chip synthesis technology gets better and more life-like every year.
I would think that human hearing is (or soon will be,) the main factor and not the nature of the source sound. Its like you're arguing that digital video will never be capable of looking like film. At a certain resolution, my eyes are incapable of the difference (not that its easy to acheive that fidelity!) Secondly, it's silly to judge anything by SlashID. I have several accounts, myself, and know people who share a single account. You're geek trash talk is weak!
http://en.wikipedia.org/wiki/Signature_bloc
to the relief of everyone in the congregation.
I can't tell you how much better a recording is than that old gas bag!
http://en.wikipedia.org/wiki/Signature_bloc
Actually I originally wrote a bit about how the new technology is more like robots playing instruments than a digital synthesizer, but /. was not letting me post it, so I tried posting again with a joke instead. I thought of artificial sermons while listening to one of the late pope's last appearances on television. He was obviously ill and incomprehensible. He should have followed the example of Ashley Simpson!
http://en.wikipedia.org/wiki/Signature_bloc