Fake Your Tutti

Thanks to Sarah Marie for pointing out this article and online quiz. [You might want to take the quiz before reading further.] It's a story about the ever-improving digital synthesizing technology that's starting to make real-sounding virtual orchestras come to life. (Even more real sounding than these guys!) How real sounding? Well, two music professionals failed to pick the fake out of a lineup of Reiner, Zinman, Norrington, and RoboOrchestra. Now, I'm proud to report that I nailed the frauds the first time and was 98% positive I was right, although I listened to each example several times before committing to a final answer. (I also correctly ID'd all the others, though I could only confirm that by submitting them in turn as "answers.")

Like most such "tests," this one has some serious flaws. First of all, all of the excerpts are sampled at 32kbs; so, not only are we not getting to hear the real orchestras live, we're hearing them in a seriously compressed format (typical iTunes MP3s are 128kbs, and many consider that a significant compromise) which tends to take some of the bloom off the sound of the instruments, real or fake. The diginstruments are, of course, just sampled versions of real instruments in the first place; so, in this setup, the real instruments are having to compete in the digital domain, and at a very substandard level. As it happens, I was also listening on some pretty crummy desktop speakers. Since Paul Henry Smith, conductor of the software symphony, has the stated goal of 'performing' with his charges live, the real test is how his pseudorchestra would compete in an actual concert hall.

This reminds me of when I bought a digital piano while in grad school. I did most of my practicing at school, but wanted something affordable at home for note-learning, etc. I actually bought the piano at the local Steinway dealer, and I spent quite a bit of time trying it out first. However, I later realized that I spent most of that time practicing it at the store with headphones on. (I'm not sure if this was particularly encouraged by the sales staff, or if I was just too shy to play in earshot of others.) When I got the "piano" home, I was definitely a bit disappointed at the way it sounded played through its internal speaker, which no doubt produces a less realistic sound than the headphone effect. The piano still did its job, and I was glad I bought it, but I've never forgotten that lesson. Since then, I've heard and played many varieties of digital pianos and, at their most impressive, what I realize is that they can sound like really good recordings of pianos.

Otherwise, the main problem with the quiz is that the examples are so brief. The full recording of the movement that they later provide is woefully unsatisfying, but that's not to say Mr. Harris won't get there eventually. So, again we come to the question of why we listen in the first place, and what sorts of meaning we draw from what we hear. I miss Joyce Hatto, and all the furor surrounding that wonderful story, because it brought up so many great questions. For example, I was recently thinking to myself that, what with the relative ease of making recordings and distributing them as MP3s, it would be great if artists just started releasing their own regular live recordings on their own websites. Eliminate the record company middlemen and only a few audiophiles would really care. (Of course, many pop groups are doing this. Actually, the Borromeo String Quartet is doing this.) Except . . . what evidence would we have that all these homemade sounds were honestly produced? Not only would we need to be wary of Hatto-ing, but Hatto's husband also revealed how easy it is to speed things up digitally.

I was thinking of this driving home last night when, again, I was listening to Maurizio Pollini abuse (in the good sense) Stravinsky's Trois mouvements de Pétrouchka. Wow, is that ever astounding playing. Here's what odd. Although I know that Pollini's technique is legendary, we still all pretty much accept that his recordings will consists of multiple takes spliced together. I doubt he could get it quite that perfect in one live take. On the other hand, I suspect we wouldn't accept it if someone learned that the music was artificially sped up to increase the virtuosity factor. The question is, why?

Now that I'm listening to sports radio again, I can't help but connect this with the whole steroids issue in baseball; interest in that topic is ramping up anew as Barry Bonds, assumed cheater, is closing in on the all-time home run record. There's a lot of complexity in that issue that deserves its own post, but most can agree on the underlying assumption that baseball players shouldn't get credit for cheating because the game is defined by the rules. Is that true of music performance as well? What are the rules? Is there a rule that says a recording needs to be something the performer could actually accomplish without the aid of an editor? In terms of cutting and pasting, we've long since OK'd that, but I'm guessing most classical music folks would still say no to "artificial tempo enhancement."

Immediately, we need to qualify that, since there's been plenty of music written that takes intentional advantage of mechanical and electronic ways of doing the impossible. Conlon Nancarrow and his player-piano concoctions come to mind first, but there are many other examples, and in the pop world no one bats an eye at such things. In fact, they're much more honest and businesslike about the end justifying the means. (Milli Vanilli is just an exception that proves the norm.) The Beatles' wildly overrated In my Life (#23!) has a famous little "harpsichord" interlude that is really just a slow electric piano solo that was electronically double-timed. Of course, in those days that also meant it came out transposed up an octave. I remember hearing a probably apocryphal story about a cello teacher at some conservatory who played some famously fast solo (Elfentanz?) at a very deliberate tempo, then doubled the playback speed to produce a viola audition tape that wowed his colleagues.

Well, it's long since become possible to doubletime a performance without changing the pitch at all. So, for all the talk that came out of Hattogate, I'm surprised so little attention was paid to the fact that some of her spectacular "performances" were not just just stolen; they were artificially made more impressive by speeding them up. As I've written before, that most listeners would think of that as cheating just proves how much we listen to most of the core classical repertoire as being about more than just the sounds we hear. There is an important athletic component that we associate with hearing the music performed.

Do we have to think about music this way? Of course not. Glenn Gould tried his best to convince us that the performer should be transparent - that the resulting sounds are all that matter. But when I hear Pollini play Stravinsky, I get a real thrill at the thought that one person is actually playing all those notes in real time. Now what if the tempo was bumped up 10%? If I didn't know about the doctoring, I'd be really astounded. If I did know - well, I don't know. My reaction would be different, but hearing it would still be pretty cool. (I'll have to try this out tonight. I have the technology. UPDATE: Done.)

Sorting all this out is really complex: our brains learn to hear certain kinds of sounds as virtuosic because they are virtuosic; then, when an artificial way of producing the sounds is developed, our brains are still left with the old associations, so the meaning we take away is a strange blend of enjoying pure sounds and interpreting them according to our understanding of human limitations. Of course, once the impossible is established as possible, it's not really impossible anymore, and our associations change. No wonder aesthetics is such a difficult field.

By the way, I happen to believe that recording technology has also changed our conception of the "possible" just by the simple process of transmitting so many astounding performances so widely. I got in trouble during my DMA oral exams for a throwaway statement in one of my papers that suggested performance standards (especially in terms of sheer accuracy) are much higher now than they would have been in previous centuries. I wasn't very well-prepared to defend that statement at the time, and I'm still not now, but I'm no less sure of it. I think that regular exposure to high standards via recordings (standards that are artificially raised by editing out mistakes) has inevitably changed the global conception of what is possible.

As meager evidence, I'll return to the four quiz recordings. The first recording that jumped out at me was Fritz Reiner's, which presumably dates from the 50's or 60's. The wind playing was simply too out-of-tune and unblended to sound like a contemporary performance. Standards are just different. ("Better" is another question altogether.) Norrington's jumped out next because of the brisk, relatively uninflected tempo. As for the sythnony, the woodwinds sustain in a hollow, straight way that just doesn't breathe right. They're too smooth. and too blended. One hardly notices the different timbre of the flutes when they come in halfway through.

Still, I don't doubt that all of that can be worked out in time. On the other hand, where I can see the value of using these virtual orchestras in theater and opera pits, I'm baffled by the suggestion in the article that Beethoven symphonies would be performed this way for live audiences. The only reason that I can imagine I'd prefer that to hearing the same thing in my living room is if the acoustical reproduction was especially fabulous - just way better than a home stereo system could deliver. But still, I believe most people go to hear a Beethoven symphony in part to witness the remarkable ensemble achievement of well-trained musicians playing together. I can easily imagine that music of the future will not be about that, though.

Surprisingly, the pop world is way ahead of the classical world on this front. From mid-career on, the Beatles were no more about live performance than Glenn Gould. If you go to this fantastic site that analyses all of their songs, it's amazing how much of the analysis has to do with the recording techniques, not just the melodies and harmonies. Still, I hope there's always a place for appreciating musicians' technical prowess as part of what makes music meaningful. And I still will dream, however hopelessly, of being able not only to hear Stravinsky played by Pollini, but to play that piece that way myself.

[I stole the post title from an old grad school paper of mine. Too good not to use again.]