When I see how a person plays music by reading musical notation, the question arises why devices could not play it instead? I do not mean playing music already recorded by people, but doing it from scratch. Wouldn't it be much more cheap and exact if musical notation was reproduced not by the hands of a musician, but by an automatic device?

  • 19
    Difference between a photograph and a portrait?
    – Tim
    Mar 11, 2018 at 7:48
  • 3
    Well, obviously cheaper is not always better, but how would "more exact" be better? There is not a one-to-one equivalence between music and notation, so what does it even mean to claim that music could be more exact if the notation were exactly interpreted?
    – user39614
    Mar 11, 2018 at 10:25
  • 17
    Devices play music all the time. All the time. Any EDM or dubstep will make heavy use of sequencing. Drum machines have been popular since the 80s, and even the most cursory look at rap or hip-hop will reveal extensive use of sequenced drum patterns. Mar 11, 2018 at 14:46
  • 15
    ... and more exact... At the risk of getting philosophical, an argument could be made that exactness is secondary in playing music; an artist's personal touch is what imbibes an interpretation with a non-mechanical uniqueness.
    – Jules
    Mar 11, 2018 at 20:41
  • 2
    I'm a musician who makes music using devices. I play both music and devices.
    – AJFaraday
    Mar 12, 2018 at 16:29

14 Answers 14


It's kinda problematic. Devices already play music. The player piano was already automatically playing music on rolls before 1910. There are pieces with musical notation written for player piano--what looks like Marc-Andre Hamelin's original manuscript for his "Circus Galop" for player piano survives today.

I suspect part of why people still like hearing interpretations of music played by humans and not just computers is precisely because of human imprecision--we lightly emphasize certain notes, vary their rhythms slightly, and even do slight tempo changes, and those all tend to make performances sound more human instead of robotic. (This also makes transcribing pieces from human-inputted MIDIs outright nightmarish--64th notes and bizarre tuplets are commonplace.)

But if you want music played by electronics, go listen to video game soundtracks. A lot of the music (especially the 8-bit stuff) does NOT involve live recordings of humans playing instruments.

  • 9
    I was peripherally involved in the task to recover the piano rolls created by George Gershwin for Larry Adler. We had to put a lot of 'fake' feel' back into those to make them suitable for a modern recording, eventually done on a Disklavier for his 80th birthday album.
    – Tetsujin
    Mar 11, 2018 at 7:52
  • 1
    Here's an example of a video game soundtrack that was not prerecorded, but played live by the computer (from MOD/S3M/IT files): youtube.com/watch?v=7MSFW8pZ-_4
    – Fax
    Mar 11, 2018 at 16:08
  • 2
    I think they were in effect cut into a moving strip as the player played; they would require later editing for control data before being mass-produced. They are, though, woefully inadequate compared to a modern midi recording, played back over something like a disklavier, or even just a good sample set. I did see the rolls at the time; though I wasn't directly involved in the actual mechanical recovery I did do some work on the Midi before it was used for the recordings. It was mentioned they were "actually made by Gershwin himself" specifically for Adler to use live.
    – Tetsujin
    Mar 11, 2018 at 17:47
  • 3
    For a more striking counterexample, Google "black MIDI" - the genre consists of songs which are physically impossible to play by hand on a typical keyboard. Typical examples may have all 88 keys firing either simultaneously or within tens or hundreds of milliseconds of one another.
    – Kevin
    Mar 12, 2018 at 2:12
  • 1
    @DavidBowling: For "Why do humans play all the music?" - they don't.
    – Kevin
    Mar 12, 2018 at 2:24

As you probably know, there are many non-human devices that play music. Dekkadeci has mentioned some historical examples; today, a typical workflow is to feed MIDI (or MIDI-like) data to a computer or synthesizer which renders the desired notes.

If you think of MIDI as a kind of musical notation, the answer to your question is just "...but devices do play notated music - it's very, very common, and yes, in many cases, it probably is cheaper!"

If you're wondering why we don't we give these devices standard notation - One reason is that historically, it would have been rather difficult to do optical music recognition. Today, however, that wouldn't seem to be such a challenge.

Another reason is that musical notation often represents a very simplified version of what a player needs to do to render a piece subjectively well. The player is usually expected to alter and add to the information that's on the score as they play it, changing timing and adding accenting information and various inflections according to their own knowledge of what would be an appropriate style to play. They will also add playing techniques that are appropriate to their own instrument according to their own personal style.

(It is also true that a human player will also produce imprecisions, and other non-deliberate playing artifacts such as breath and string noises, that may be considered attractive.)

I imagine it would be possible (if not easy) to get computers to also do that stylistic interpretation, but for whatever reason, commercial products that do this aren't common - typically, whatever expression you want to add to your performance has to be programmed in explicitly. And this can be a lot of work - it is often easier (and, therefore, actually cheaper) to give the piece to a skilled player of the physical instrument who has a lifetime's experience of playing pieces in a stylistically-appropriate way.

I guess it's that aspect that is the key to your question - will machines that can interpret simplified representations of music, as a human player does when they read a score, ever be common? I'm sure it has been done, and I am fairly sure it would be possible to do well, but I'm not sure it fits in with the workflows of how music is commonly made live and in the studio - when we need interpretation, we tend to rely on a human to provide that.

There are also a couple of reasons related to human behaviour as to why we don't let the machines take over. One is that people rather like to see an attractive/famous/well-dressed young man or lady (or group thereof) playing a piece. Or their kids, if it's the school orchestra - there are lots of social reasons why people want to feel that the music is a communication from one human to another.

The other thing is - playing music is usually enjoyable - so why let the machines have all the fun?

  • 4
    With respect to your point about programming in playing nuances, that's a huge part of making a vocaloid performance, and those are fairly popular as far as non human performances go.
    – Tama
    Mar 11, 2018 at 22:22

Even if “more exact” were clearly a desirable goal – how would you even define “exact”? Musical notation is not really a mathematically well-defined syntax, although it has many mathematical aspects. Perhaps the most straightforward is rhythm / metre: you might state something like

⁴⁄₄ time means, you start with some unit time span (a bar) and divide is into four equal portions, the beats. Each beat can further be subdivided into a power of two of equal quavers / semiquavers / etc., or indeed any integer number of equal tuplets.

But does it? That description certainly gets pretty close when you want to describe rock/pop, but musical notation is much older and through most of its history – as well as for performances today of romantic, classical, baroque or renaissance music – that's not quite how it works. In classical music, ⁴⁄₄ time rather just means you group four elements to a bar, not really that these beats are metronomically equal. In fact it would widely be considered incorrect if you play, say, all three beats of a Waltz in the same way. The 1 should not only be stronger, it should also usually come a bit later than a metronome would suggest. How much exactly? It's not defined, up to the conductor / soloist / concert master to judge.

It gets even more tricky when you consider pitch. Although many consider the 12-edo tuning (as found on a digital keyboard or well-tuned guitar) to be “the exact standard”, this too is a relatively recent invention, and unlike with rhythm it shouldn't even be considered the mathematical ideal: the physical underpinning for the phenomenon of harmony is resonance between overtones, and that is achieved in just intonation. 12-edo offers decent yet audibly imperfect approximations to just intonation, and good singers and string players can thus reliably out-intonate normal digital keyboards / sequencers. You could now say, well, surely we can tune a digital synthesizer in just intonation? Well, you could, but here we run again into the ambiguity problem: most music does not live in one single definitive just-intonation space. Rather, you have some chords that have a clear relation of JI frequency, but the relation between different chords is often not so clear – there may be two or three different ways they can be derived as JI. Trying to fix every interval to some exact integer ratio will often lead to funny problems like comma pumps, so in practice you need to give up making everything perfect. In fact, it's arguably desirable to make some chords a bit rougher than they would be in JI, so as to increase dissonant tension of a dominant and make the resolved tonic sound all the more harmonious by comparison. Pau Casals was (in)famous for this, most other players tend to keep rather closer to JI or indeed to 12-edo. Once again, it's up to the interpreter to settle what's “perfect”.

So, in order for a machine to actually be able to play “exact”, you'd first need to define what kind of perfection you're aiming for, and in many cases the only way to do that would be to measure timing and pitches of the best human players. But then, your machine can never be more exact than these human players.

  • There are other actual benefits to computer-played music, which is why e.g. games use it. File size for individual songs is a lot smaller (compare to gigantic FLAC). Playback is limited by the limits of the current equipment being used, not the recording equipment (compare to lossy MP3s). Mar 16, 2018 at 18:33

Individual interpretations tend to me something other than the score they are working from. I am pretty good at working from scores, but it often takes several double takes until I find myself making a rendition that I find compelling and making sense to me. If you wrote down the intermediate stages, they'd still match and produce the score.

Now I am playing accordion. It is sort of interesting that those differences matter more when the instrument is registered sparsely, where the tonal changes from bellows pressure don't average out turning the bellows into a mere volume control, and I can channel a flute or harmonica or violin.

As a similar observation, Midi expanders tend to have a much harder time producing a reasonable solo violin than an orchestral violin section. Articulation is notated sparingly and yet is a very important part of interpretation, and with solo instruments, articulation, breathing pressure, bow technique, in connection with subtle note groupings and accents are an important part of the performance.

For backdrops, automated music like drum computers, arrangers, or Midi tracks straight out of notation programs may actually be a convenient substitute. Substituting solo instruments is more detrimental to the appeal of the end product.


There is a lot more to music than just playing notes.

  • A lot of music tells stories: In order to tell the story, you need to understand it, so you'd need to teach the device what to see in the story, how it connects, what the large ideas are.
  • Even if you don't have a story per se, there are still ideas behind good music. These ideas are expressed not only in notes, but how they are played and in particular, how groups of notes are played. This is called phrasing and one of the most important aspects of listening to music. When most people listen to different interpretations, they'll often think that one is not as good as the other, or maybe one is even boring. Often, that has to do with musical phrasing. Phrasing is not just a technical aspect, but it needs musical comprehension: You need to know what the phrases in a piece are and what they mean together, i.e. you need to understand what's going on.
  • In addition, different pieces have different moods, which will also affect phrasing. There are certain things you can do to help convey the moods, but you need to understand what it is. It's not just as simple as playing in minor when you are sad or playing in major when you are happy. There might even be irony involved.
  • Live music is (heavily) affected by the environment and the feedback from the environment to the music is one of the reasons most people enjoy live music more than just playing an mp3. Why? Because the environment affects everybody's mood, which will make people play differently. If you regularly go dancing, you might have experienced that effect - that's why you have a DJ: to adjust the music to the crowd, because it can't do it on its own. However, most people agree that live music is even better than the best DJ, because the musicians can do more to help the music adjust to the crowd than the DJ can.

If you aren't a professional musician, you won't hear a lot of these nuances and you might not even feel many. But the more you listen, the more you learn - and the less you'll be satisfied with machine-only renditions.

Everything that I said above are things that can be learned and are related to technique. I'm not saying that a machine can't do that, I claim that machines can't do it yet. Maybe in a couple of years, maybe in thirty years we'll have AI making music in the style of whomever you want and new styles will emerge from that.

But rest assured, most music will still be made by humans.

Why? Because most music is at its core a social endeavour. We feel that it tells us something about us or about the world around us, we think that it connects us to others or that we can understand these people. Young people cry and faint when they see their favourite star. Nobody would do that for a machine. So even if there wouldn't be the technical problems I pointed out above - it would be cheaper, but nobody would buy it. There are some things you just don't get cheap.

  • I'm afraid most of that is faith, not fact. (And I AM a professional musician., and I don't like it either.)
    – Laurence
    Mar 13, 2018 at 11:55
  • What do you mean by "most of it"? The technical aspect (phrasing, mood, etc.)? I'd say that's a fact but since I'm no musicologist, I have only anecdotal data to back it up. The last paragraph? Well, I'd say there are a lot of examples for that, e.g. the Japanese group AKB48 whose success is based precisely on the fact that the interpreters are human - the music is... less than noteworthy.
    – Martin
    Mar 13, 2018 at 18:27

In addition to all the other answers: not only are people rather good at making music, but many people rather enjoy making music themselves. I don't really care if someone can reproduce my music more "precisely" with software: I play for myself anyway, and also for others. I have nothing against music interpreted by silicon, but I'm a human, and so are my listeners, it's fun, so what is the advantage? Let computers play music for other computers.

  • 1
    people rather enjoy making music themselves Great answer! If I just wanted the 'best' music, I would never play a note - I would put on recordings of the greats - people who I could never reach in 1000 years of practicing. Thing is, I like to play.
    – Stinkfoot
    Mar 13, 2018 at 14:45
  • 1
    @Stinkfoot- that's exactly it. Mar 20, 2018 at 17:13

(This answer encompasses what has been said in other great answers about the human factor regarding playing music. It's simply looking beyond the details and generalizing those concepts. It takes more of a 'philosophical/psycological' angle than a 'musical' angle.)

Why do people, not devices, play music?

  • Why do people pair off with other people, and not their automobiles, even though automobiles get around much better than people do?
  • Why do most people feel lonely without other people, even if they are surrounded by 'intelligent' machines that are much 'smarter' and more reliable than other people?
  • Why do people (hopefully..) value their friends more than their smart phones, although they are constantly using and attending to their phones?
  • What do most people prefer playing chess with a human opponent rather than a computer, even though the computer never gets tired, never cheats and is always in the mood for a game?
  • Why do so many parties and venues use a live band, instead of a DJ or a juke box? (Maybe that's part of your question...)
  • Why do musicians prefer to play with other musicians, rather than with a click track or MP3's?
  • ... We could continue adding countless items to this list.

The answer to all of these questions, including yours, is the same:

Humans, being Human, relate better to other humans and their actions, than they do to machines; people tend to prefer People over machines for their interactions and compansionship. Machines are... Machines, not people. They do not act, react or think like people do.

The particulars of why people are that way with respect to music, (and all the rest of their activities) can be broken down and explained with myriads of details - some of those details are well articulated in the posted answers. Still, this fundamental point transcends any particular details, and is perhaps impossible to completely and accurately quantify or detail at all, although virtually all the answers have highlighted different aspects - it's fundamentally intangible.

Similar behavior is found in most other species, and perhaps even in certain inanimate objects - it's the nature of things.


It's absolutely doable and it is done. There is plenty of software that lets you enter music in musical notation and that will play the music via some device connected through MiDi. Usually the device is some synthesiser; it is entirely possible to connect a piano, or a church organ, or many other devices, just a mechanical problem. And reading musical notations from sheet music is not that difficult.

Why this will not produce particularly nice music is because of the limitations of musical notations. If you give some sheet music to a pianist, then after some practice he or she will not be playing what the sheet music says. They will play this note a bit louder, this note a bit faster, and there is a certain ability to change the sound of each note (not so much on a piano, a lot more on other instruments). And that difference between played music and sheet music is intentional (for a good musician) and improves the whole listening experience.

The other question is: Why? If there is a concert of classical music in some concert hall, would I pay money to listen to music (a) played by 50 real musicians, (b) played by a computer and 50 musical devices, or (c) recorded on a CD and played back on some good speakers?


A skilled musician playing a typical instrument will manipulate it in many subtle ways in an effort to achieve the composer's intended musical result. While it would be possible for someone at a computer to manually specify all those manipulations, it might take hours to make all the adjustments necessary to produce a minute of perfect-sounding music. By contrast, a musician or group of musicians that is being recorded could produce a minute's worth of usable music in only 60 seconds.

  • Another important aspect. Perhaps it's just superstition on my part, but there's also something special to me about a live performance, where the music is being made right here and now. Mar 21, 2018 at 10:12
  • @ScottWallace: There definitely can be, particularly with a group that meets regularly. In a very real sense, the people can form themselves together as an "instrument" which is played by the leader/director. I'm not sure to what extent professional groups reach that point with pieces that they learn for the purpose of making a one-off recording, and may never play again, versus groups that add pieces to a regular repertoire.
    – supercat
    Mar 21, 2018 at 14:38
  • yep. Unfortunately, we all know examples of live concerts where it's obvious that the musicians just got together two weeks before and rehearsed maybe twice. At that point, the immediacy of the live performance becomes embarrassing. Mar 22, 2018 at 11:08
  • @ScottWallace: That ties in with a distinction I heard someone make: the distinction between rehearsing and practicing. Rehearsal is focused on learning a particular group, while practicing is focused on more general aspects of performance. Some groups can decide to play a piece of music on the spot without any rehearsal and nail it, if all the members have enough practice working together.
    – supercat
    Mar 22, 2018 at 14:41

devices do fine with time accuracy and hitting all the notes.

but a human plays eeeach note at a particular volume. and probably a little different each time.

the only way to give that selection of volumes for eeeevery note to the computer is to record it live. (or step entry, whatever).

that selection of individual "velocities" is what makes a human played song sound amazing. Also, a human will vary the tempo just a little in places.

listen to techno and EDM - THAT is what a computer can play. It gets boring fast. listen to some classical piano or 70s songs where humans actually played the music. you can tell that humans do it better. and always will.


Recorded music HAS overtaken live performance to an enormous extent. So far there's usually SOME human input and the computers are used to correct and enhance the original performance. But I'm afraid the people who complacently insist that technology will never equal a human performance by a great artiste will eventually be feeling as silly as those who insisted a chess computer would never beat a Grandmaster.

You may consider this exciting, you may consider it sad and terrifying.

  • 5
    The huge difference between chess and music is that the former has clear, mathematically stateable winning conditions. Music doesn't. So it will always be up to opinion whether a computer is playin better than the best human, even though for any given criterion what's considered perfect, computers will increasingly effortlessly win. Mar 12, 2018 at 1:29
  • 1
    @leftaroundabout - Excellent point. StarTrek's Commander Data, an android, was programmed to emulate perfectly the styles of 300 great violinists. But, he did not have his own style because he was not human - something which he often lamented.
    – Stinkfoot
    Mar 13, 2018 at 8:24
  • @Stinkfoot - yep. But perhaps Commander Data 2.0 might develop his own style. I don't know if we'll ever get that far, or if we should, but I don't see any theoretical hindrances to making hardware perform as well as wetware. Mar 21, 2018 at 10:16
  • @ScottWallace - Not a believer in "wet-ware" - we're human, not 'wet' computers. See the writings of the physicist Sr. Roger Penrose - Emperor's New Mind, Shadows of the Mind...
    – Stinkfoot
    Mar 21, 2018 at 18:04
  • 1
    @Stinkfoot - perhaps we'll have to agree to disagree about this. It's certainly true at the moment that there's no comparison between humans and anything computers can do in the way of, say, improvising a jazz riff. And computers might never be able to compete with us. In any case, I'm all in favor of human participation in music. Mar 22, 2018 at 14:07

Written music is not, and has never been, more than an approximation of the emotions the composer wished to convey. Different musicians will all play / perform a piece at least slightly differently, and you as a listener will enjoy some of those more or less than others. Without the emotional content provided by the both performer (and the listener), those vibrations would not be music, but merely noise.

  • those vibrations would not be music, but merely noise - there is plenty of synthetic music that is not 'noise'. It's just not usually as satisfying as hand-crafted music.
    – Stinkfoot
    Mar 13, 2018 at 16:05

Automation is replacing humans in repetitive tasks every day, from calculations (use Excel instead!) to washing cars to getting recommendations for a good pizza place. Reading the notes is a repetitive task: why are humans wasting their creative potential and common sense struggling to read notes and play back with precision what somebody else told them to play back?

Good question! It's because music is a form of communication, and it's no fun communicating with a robot.

I just wrote the following definition for music, which I'm now peddling here on music.stackexchange.com:

Music is when a person uses sound to communicate ideas other than language.

Perhaps you are questioning the value in learning to reproduce written music on your instrument with accuracy and precision. Don't worry! ...there is much value in this: without accuracy and precision on your instrument, you cannot clearly communicate musical ideas either as a soloist or in an ensemble.

If you cannot communicate the ideas clearly, then you can't make the music. And that takes you out of the equation!

Learning an instrument is hard, like learning to paint, ride a bike, speak, or walk is hard. But there is great reward in using your instrument musically once you overcome the initial hurdles. Don't give up!


My answer is short and simple. People have the ability to imagine and hear in their mind things that we can refer to as music. Things that have no mind can't do this.

Not the answer you're looking for? Browse other questions tagged or ask your own question.