Veni, MIDI, Non Vici

I’m in a quandary. Oh, not much of a quandary, just a little composer’s mini-quandary.  In the last year, performers have started asking me for MIDI mockups of pieces I’ve written for them. I don’t recall this happening before. Of course I provide MIDI versions in the case of the occasional acoustic microtonal piece, in which the performers need to learn pitches they’re not used to hearing, and my microtonal music consists only of MIDI “versions,” in which cases I humanize the hell out of the file and avoid instrumental sounds that don’t sample well (no arco violins, ever). But I have misgivings doing it for “normal” music.

Personally, I don’t mind listening to MIDI versions of pieces not yet performed. I have enough performing and rehearsing experience that I feel I can “hear through” the stiff MIDI limitations and imagine what the piece will actually sound like. It’s an aid, and you have to know how to use the aid and not mistake it for the reality. But I’ve also had enough bad experiences playing MIDI versions for people – people who didn’t seem to possess that ability, and who reacted with a visceral dislike to the piece based on its synthesized version – that I avoid playing MIDI versions for others. 

Further than that, while it’s one thing to listen to a MIDI file to get a sense of a colleague’s new piece, it strikes me that to listen to a piece that way with the intention of performing it later is an entirely different thing. I’ve had some bad experiences with it. Sibelius (the software) playback has trouble with fermatas, which sometimes get applied to notes they weren’t intended for, and it doesn’t always portray arpeggiated chords or grace notes elegantly; the result being that I have sometimes had to go back and remedially convince the performer that I meant what I wrote in the score, not what he heard on the MIDI file. Of course I could import a Sibelius file into Digitial Performer and sculpt every note, but that’s an awful lot of work for something that isn’t a final product, just a temporary convenience, and the results are still imperfect. Even when such details aren’t a factor, I’m uneasy with the idea that a performer’s first audio experience of a piece will be through a stiffly metronomic version with no nuance. There’s a wonderful process that happens in rehearsal as the players slog through notes whose interrelatedness is still a mystery to them, and then suddenly they hear other parts correctly played in hocket with their own, and everything clicks, and the music emerges from chaos. Then they create the piece from the notation, and it has personality, rather than trying, however unconsciously, to replicate a lousy artifact they heard an mp3 of. 

But that process takes time, and when an ensemble is short of time, they ask for a MIDI file to speed the process up. As I’ve written here before, “Efficiency in the pursuit of music-making is no virtue.” But I hate to turn down a request from people enthusiastic about playing my music. Perhaps I’m being too sensitive and old-fashioned about it, and this is the way things are done these days, and I’m curious what other composers do and how they feel about it. 


  1. says

    Hi Kyle,
    I’ve struggled with this as well. I don’t generally give my band MIDI realizations beforehand unless absolutely necessary (and I’ve learned through experience which compositions are likely to fall into that category.) But would it be possible for you send them the MIDI version after they have their first rehearsal? Wouldn’t that allow for at least some of the experience of trying to create the piece from the notation first, and then throwing them an assist once they have gotten started?
    I just tried this at yesterday’s rehearsal with the new piece I’m working on with my band — hack through it in rehearsal once, then send out the MIDI demo so people can, if they wish, check it out before the next rehearsal. (Some performers have the same reservations about listening to MIDI realizations as you do.) We’ll see if it helps.
    KG replies: I guess if you have your own group, as I once hoped to back in the ’80s, there are different things you can try. It does sound like an improvement.

  2. says

    There’s always been a lot of pooh-poohing over MIDI mock-ups, yet at the same time the frequency they’ve been asked for does seem to be increasing. I ended up making mock-ups for my old professor — one a string quartet, the other a viola concerto — purely to snag performances (and it worked, too). I think part of it is the move away from everyone having all that intensive score-reading training that used to be the norm; there’s also the the fact that a strange forms, techniques or tunings can be hard for a performer to wrap their head around just from the score.
    You know how I feel about MIDI in my own work, since it’s on your PostClassic Radio as we speak (& a big thanks for that). But then I work back-assward: composing and perfecting the MIDI first, and leaving the score extraction to some brave soul in the future. In some way, I suppose I compose exclusively electronic music, that just happens to sometimes sound or act acoustic.
    KG replies: I’m not surprised to hear that you use MIDI in your music, but much of it doesn’t sound like MIDI, since the sounds are so distinct and used sparingly.

  3. says

    I’m surprised, actually, that performers would ask for a MIDI file, given how often people diss MIDI files (or more commonly, MP3 files generated from a MIDI-based realization via Finale, Sibelius or whatever). Most of my stuff falls into this category. Some of the sound files are pretty adequate, and others are…well…not so adequate. As you mention, strings are really really hard to sample well, and as a former violinist, I have yet to hear a sampled solo violin that wasn’t obviously sampled. Keyboards, on the other hand, work pretty well, as does flute and a few other instruments.

    I do, however, take the time (and often, an inordinate amount of time) to make sure the final sound output is good enough, since in many cases, that’s all anyone will ever know of my piece. For the most part, the notation part of Finale is pretty standard; what’s tough is the Save as Audio part. I’m glad to see Sibelius also has its quirks in that regard.

    I’ve also found it invaluble to compare live performances with the Finale-generated versions. Even though the Finale output is metrically and note perfect, the live performances (not surprisingly) are just preferable, period. That said, I also think that what certain virtuosi like Steve Layton can do with sampled instruments (like ) shows that with the proper tools and talent, sampled instruments can actually work quite well in conveying the music.

  4. says

    An Australian newspaper had a competition a few years back for modern versions of “Veni, Vidi, Vici”. As I recall, the winning entries were:
    Veni, Vidi, Video — I came, I saw, I took photos.
    Veni, Vidi, Versace — I came, I saw, I looked fabulous.

  5. says

    usually, i fall into the anti-midi camp, but recently i took the time to make a midi mock up of a piece that had been performed before, but never felt like it had enough rehearsal — as both the composer of the piece and a performer in it. it was a huge help to me, but one of the other performers found it more confusing. so what can you do? i’m sold on it, actually, and will probably make more in the future.

  6. says

    Obviously this is a situation where different solutions are more appropriate to different kinds of music and different attitudes, but I’m certainly generally in favor of MIDI mockups.
    It’s true that listening to a MIDI mockup requires imagination and extrapolation from the listener, so I rarely show the mockups to non-professionals, but I always (which admittedly for me has been a pretty small number of times) offer performers a mockup. One of my main reasons is that I tend to write a lot of rhythms which are ugly on the page but make intuitive sense to the ear, and I hope that by providing a mockup I can reduce the amount of time a player has to spend learning notes and rhythms and free them up for fine-tuning and interpretation. I don’t worry much about the stiffness of the performance because I tend to trust the performers to be able to make it musical. I don’t know to what extent performers actually make use of those mockups, though.
    It’s also easier for me to prepare a MIDI mockup than it is for a lot of people because I compose in Cakewalk rather than in a notation program. I prefer the tools available in the studio environment, both in terms of moving the notes around and generating quasi-realistic sounds. In my experience MIDI mockups generated from Finale/Sibelius sound terrible, but because I have more control over some key things in Cakewalk I can get pretty good results with only marginally more effort.
    My basic strategy is as follows:
    My main sample library is Garritan Personal Orchestra, which sounds great in almost all instruments except solo strings, which are always a disappointment. (I’ll sometimes use the ensemble string patches as a standin for solo instruments just because they’re more musical.) I usually apply some compression to the GPO output, and I often split the instrument assignments between a few differnt instances of GPO so that the compression doesn’t result sudden changes in volume of the sustained instruments when pianos or percussion play–I usually assign percussive instruments to one instance and sustaining instruments to another. I always add some reverb, which makes an _enormous_ difference in realism (one of the problems with most Finale/Sibelius generated mockups is the lack of reverb). Setting the panning is also critical. Once reverb has been added the panning settings can be quite wide, giving room for the instruments to breathe without sounding artificial.
    For the MIDI data itself, I draw any broad dynamics in with either a velociy envelope or by drawing the velocities by hand, and create any significant accents by editing the velocities. (I would guess that Finale/Sibelius would create these velocity adjustments automatically in conjunction with markings in the score, so a MIDI file exported from them and imported into your DAW of choice would require less of this hand-editing.) Then I use a quantization plugin to randomize velocities by a small amount (I’ve found 6 to 8 out of 127 is usually about right for adding some humaniation) and I sometimes randomize the rhythms slightly as well. I also globally scale the velocities of each instrument by a percentage, since the ideal velocity for some patches is different from others. This is especially crucial for pianos–the best normal velocity for most instruments is around 100, but for piano it’s much less–I generally scale the velocities down to around 60%. Leaving piano velocities too high is one of the most common mistakes that results in artificial-sounding performances even with the best sample libraries.
    I don’t have any MIDI mockups on my website other than solo piano, but “Two Pages for Lee Hyla,” “Ex Nihilo,” “Distance Over Time,” and “Scherzophrenia” are all MIDI. They’re imperfect, but I do think the performances are musical.
    Kyle, if you want to send me a MIDI file of something I can try making a mockup of it and we can see whether you’re any more comfortable with the results I get than what you’ve been getting.
    KG replies: Thanks, Galen, but the piece I really needed it for was my piano concerto, the MIDI of which was pretty horrible (not least because the orchestra was mostly brass, which MIDI handles *terribly*, and also because the piano part had no fluidity) – and that piece has already been performed.
    On second thought, if you’re really sitting around with too much time on your hands, I’ll send you “Uranus” from The Planets.

  7. says

    On a releated note, another helpful tool (after somebody picks up on the piece, of course) is to record the rehearsal. I remember back in grad school, having a piece for three violins. When we got together for the rehearsal, I just happened to bring along a small tape deck to capture it. Somewhere in the middle of the toil, I played back a little, and I’ll never forget one of the violinists commenting “Oh, so that’s what it all sounds like!” It was my first realization that very many performers ears might not be hearing the big picture while they sweat the details of their own part. That little bit of recording made the rest of the rehearsal go quite a bit smoother.
    KG replies: An infinitely preferable idea, with no downside.

  8. says

    Hmm. I’ve never been asked for a midi/.mp3 mockup of a piece that I can recall. As a performer, I’ve frequently been offered them by composers – in fact I’ve got one I’m supposed to be listening to at the moment – but I find I very rarely listen to them, I prefer to hack away at the notation.

  9. says

    I think most performers who have worked with MIDI realizations understand the limitations and aren’t afraid to insert their own interpretations and/or personalities into the live performance.

    If a performer asks you specifically for a MIDI realization, I see no reason not to give one to him or her. I like to think of it as another kind of performance practice that’s been evolving since MIDI realizations became possible. A MIDI realization can’t stand in for the real thing, but then again neither can a score.
    KG replies: Your last sentence is God’s own truth. I agree that good performers are able to hear past the stiffness and infuse the piece with their own personalities and musical playing – I’ve seen it happen. But I still think that there’s something about relying on MIDI that offers a specious illusion of familiarity with the music, and accustoms each player to the musical surface, decreasing reliance on the more visceral cues he gets from other players in the performance situation. It seems to me, based on an admittedly small number of rehearsal situations, that the reliance on MIDI rather than hearing the other players makes it a little more difficult for the sound of the piece to *gel*. Or at least, the decreased rehearsal time for which the MIDI version compensates has that effect. (Sigh.) Maybe not, or not always.

  10. David DeMaris says

    There are a couple of options that may be worth exploring. I’d particularly recommend trying to get your hands on SuperConductor which instantiates the theories of Manfred Clynes on the importances of temporal microscale adjustment s for conveying emotional content in music. (I’m sure we could have a long discussion on this if and when you dive in). I don’t have any personal experience with it, but would definitely get it if I were doing more score based work. The demos on the site speak for themselves (but the player apparently is Windows only). I think it will make a guess at the right parameters for any music, but you can also teach it your “style”. The demos show you the result on the classical canon. It comes with its own audio engine so that volume and articulation, including vibrato are adaptively modified, not just note timing. Jeff Harrington, whom you appear to know by his comments on, uses it extensively and given your proximity might be able to give a test drive or tweak a piece for you.
    Less radical is the synthesis afforded by plugin Synful Orchestra, which does a lot with note sounds and articulation but not so much with timing.
    KG replies: Very interesting. Only Mac here, though.

  11. says

    Garritan’s orchestral brass is gorgeous, actually, although it’s true that an awful lot of synthesized brass is pretty bad.
    I dont’ know about “too much time on my hands”. . . :) but I’d be glad to take a look at “Uranus.”

  12. mclaren says

    Some of us consider our MIDI files the definitive version of the composition — particularly those of us who compose for an orchestra of superhumans playing tunings no live orchestra will ever be able to produce (41 note per octave non-just non-equal tuning based on the inharmonic vibrational modes of a sphere, anyone?) and rhythms far beyond the capacity of any orchestra or live ensemble to recreate (melodies in tempo ratios 30:31:32:33:35:37:39:41 with embedded 5- and 7-tuplets with triplets inside the tuplets, anyone?).
    This helps explain why Western musical notation is blowing apart and increasingly vanishing from postclassical musical culture.
    You remark:
    I have sometimes had to go back and remedially convince the performer that I meant what I wrote in the score, not what he heard on the MIDI file. Of course I could import a Sibelius file into Digitial Performer and sculpt every note, but that’s an awful lot of work for something that isn’t a final product, just a temporary convenience, and the results are still imperfect. Even when such details aren’t a factor, I’m uneasy with the idea that a performer’s first audio experience of a piece will be through a stiffly metronomic version with no nuance.
    At the very least, every composer anywhere close to the cutting edge today needs at least three different versions of each score: [1] a common practice classical score (if possible) to enable analysis, simply because traditional musical notation compresses so much info so conveniently. For understanding the piece, nothing beats a common practice notation score. Even if it’s replete with xenharmonic accidentals, you can tell where the triads are and what the rhythms are at a glance. Can’t do that with MIDI or Csound. [2] A MIDI score to serve as a guide, particularly if the pitches or the rhythms depart markedly from western common practice. Live performers usually haven’t heard, say, 79-limit just intonation with 37 pitches in the non-octave ratio of 3:1, so they need a sonic blueprint however crude to assure ’em they’re in the right ballpark when they rehearse; and [3] a fully-nuanced MIDI score complete with expression and rubato, accelerando, etc., because (let’s be honest) most compositions don’t get performed, and if they’re performed, it’s usually by an ensemble that doesn’t have enough rehearsal time and will make plenty of mistakes and introduce enough slop into the timing that it becomes problematic. The fully nuanced MIDI version substitutes for the error-filled live performance and represents the definitive version of the piece…at least, if the composition ventures anywhere in the ballpark of the edge of contemporary music.
    Twelve Tone Systems made a program called IN CONCERT that lets you tap in a tempo and control the playback of a MIDI file. How hard you press the key on a MIDI synth (or how hard you hit a MIDI drum pad — could be any MIDI input device) continuously the overall MIDI volume and so makes the music louder; how fast you tap tempo controls the overall playback tempo. In effect, IN CONCERT lets you conduct MIDI pieces in real time. You can save the resulting MIDI file with expression, rubato, etc. IN CONCERT runs on both Windows and the Mac.
    Ideally, contemporary composers should also produce a Csound score (great for showing the actual Hz frequencies of those non-Western pitches) and a piano roll score (lets you see multiple overlapping tempo streams at a glance, particularly useful if you’ve also got controlled progressive accelerandi and decelerandi going on at the same time).
    The bias against MIDI performances represents a vestigial throwback to the era when live musicians were cheap and electronics were expensive. Today, that equation has reversed, and as electronics gets cheaper and more powerful, the economic effect is inevitably that live performances will grow exponentially more costly relative to any electronic realization. This means that live performances of all music will grow increasingly rare relative to the number of electronic realizations, at the same time as the electronic realizations improve in quality. Indeed, this is already happening. Ask yourself: how does the average audience member discover that great new contemporary composers? Via CD? A live concert s/he attended? Or by downloading (or streaming) an mp3 over the web? Chances are it was the latter.
    If you doubt this, posit the effects of our upcoming $20 per gallon gasoline on concert attendance. At some point in the near future, electronic realizations of contemporary serious music will become the de facto way new music is heard (if indeed it’s not already the case), and people skilled at making the composer’s raw MIDI versions of compositions “sing” by adding nuance will become valued and respected members of the serious contemporary music community.
    Add to that a wealth of new live MIDI performance devices like Yamaha’s Tenori-on and the monome, and you’ve got a recipe for an intensely interesting 21st century concert scene — live, but not acoustic; nuanced, but not performed by individual human performers (Synful orchestral truly is amazing, but so is the Tassman physical modelling percussion softsynth). The Princeton Laoptop Orchestra is only the beginning: CCRMA’s 2008 Summer tutorial program includes a session called “Cell Phone Orchestra.”

  13. says

    I am too often heartbroken by MIDI reactions even of musicians whose experience/ability I hold in great esteem. But I just don’t have good audio recordings to show in their place, and most musicians can’t look at a score and make a reasonable “listening experience” out of it. I think it’s important enough for people to hear my music at all. So I put them up as MIDI files on my website.
    I fear what kind of damage this is doing to my writing.
    KG replies: To quote a recent politician, I feel your pain.