https://www.brain.fm/ does this. I enjoy using it while working sometimes and I think it works very well as believable background electronic music.

I believe that if you look at the work done at IRCAM since 1977, you'll find a lot on generating music—from scratch, or from interacting with musician, or from gestures… (Max/MSP was also invented there before it was a Cycling74 product.)

So there is research on generating music, but like a lot of cool research, nobody is making an effort to turn it into products.

AAA game studios have 100s of artists toiling away at Maya and ZBrush, so I guess that means there isn't a working generator for computer graphics either ;).

In this case the author is using math to paint shapes and fill in colors, but the wider scene composition is authored. The equivalent of this for music would be hardcoding music composition/sequencing using code and writing procedural synths using math (like http://www.shadertoy.com music shader!), not a server somewhere generating music with no human involved.

Entirely procedural music is a genre though, as far as procedural music being "believable" (I guess aesthetically pleasing?) for you it probably depends on the right artists coming along and developing a sound you like, moreso than it not being possible.

There was http://www.abundant-music.com/ and although it doesn't seem to work now, you can look on YouTube for some examples of what it can generate. It's massively configurable.

Maybe it's that music matters and even instrumental music contains a similar level of content and meaning as speech.

The minute we realise music isn't the authentic voice of another human being we reject it.

Even my enjoyment of arguably mechanical music like some of the early minimalist stuff - where to a degree processes determine the structure and unfolding of the work - is mediated through my awareness of the authorial intent of the composer. (I'm talking more about Piano Phase than Pendulum Music here - the latter is slightly too devoid of human intervention to be 'real music' for me)

How would you feel about a novel generator? Why is that so different?

Are you familiar with modern electronic dance music? There’s definitely no “authentic voice of another human being” in regular trance music, for example. It’s hours upon hours of repetitive structural patterns overlaid on similar timbres. Now, I’m not a very big fan of trance music, but it would be something, and I feel like the gap between that and more interesting music like IDM should be quantitative (more code to produce more complex patterns over more varied timbres), not qualitative.

Modern "EDM" producers actually receive a lot of criticism for these practices from both critics/academics and laypeople. I think the massive commercial success and general enjoyment of this music actually comes from the spectacle of the "live" events, and so as an extension even just listening to it at home, the spectacle is implicit and can be experienced vicariously to some degree. On top of this, this music as a product is obviously extremely easy to produce in comparison to other genres/processes and is thus lucrative to profiteers.

Another reason why such utterly cheap production habits have been wildly accepted is because of the already established history of house music and techno, which was originally revolutionary, because those sounds and the contexts for which they were experienced ironically represented newfound human feelings of liberation, from a youth movement to escape a bankrupt Detroit to the fallen Berlin wall. Techno was a way for people to grab the rising technocracy by the horns so to speak and make something specifically human out of it.

To me, this just indicates that you aren't familiar with what the human element of modern electronic music is. I definitely understand where you're coming from with trance, I'm not the genre's biggest fan either, but there are certainly methods and techniques of production that give artists unique characteristics. "EDM" that doesn't sound like it was produced by an artist with some sort of vision or goal in mind is simply bad EDM. It's extremely obvious when someone releases a by-the-numbers boring dance track, it's the same thing as the awful pop that plays on the radio.

Anyway, my point is: even music with "repetitive structural patterns overlaid on similar timbres" cannot be replicated with any genuineness by computers. Not yet, and probably not for a very long time. We, as people, can sense the insincerity. Now, if the goal is literally to have repetitive background noise with no real melody or structure then sure we can certainly already do that. But I also would not call that music in the true sense of the word.

Ah, good find of the github link! I imported it into glitch so I could see it run: https://honey-climb.glitch.me/

Original dev here. I'm actually working on something like this, too. :)