Automatic music : is it possible ?

[jancivil]

“In terms of determining which scale(s) a piece of music is in, I would imagine a probabilistic approach over time.”

Why? What does “over time” do to change the fact of a scale in some music?
Give a musician any tonal music (or modal) and the scale is apparent. It’s not typically more problematic than a determination of key if tonal or harmonic. It’s not a guess.
It’s not quantum theory, a person’s observation or perception changes nothing about what it is.

To get to the meaning of the OP, I looked at examples on their youtube. EG: (Charlie Parker) Donna Lee as though redux, an imaginative constructive approach with interesting results, but there’s nothing groundbreaking in it nor does the ‘problem’ of scales to its harmonic underpinning demand this much intellectual prowess or an especially advanced solution. It strikes me as impressive in itself but uninspiring compared to music in real time by ear.

I don’t think it’s wrong, the results show expertise in the usual ways, and one supposes this was basically a lot of fun for them to do.
As far as determinism by a mathematician being particularly suited to note choices in an already determined harmony - and here the form is a function of the tune/its harmonies or the point of the choice of tune is, ah, unclear - it’s barking up a tree no one has to climb.

[music-engine]

@Jancivil

Your recitals are interesting, and show what music should be for a musician.
But the problem of harmony is not as simple and intuitive as you say.
Of course, if we remain in the schemes of tonal music it is true that the scale is easily determined. But in the music I put on youtube there is a very complex harmonic research.
If, for example, I take the harmonic sequence:

Fm / Em / C7 / C#Maj7

what are the scale you'd use?

The main problem in understanding harmonic structures is to look for a segmentation, which seems trivial but is not. I have proposed in France a theoretical approach to the segmentation of harmonic structures. Unfortunately the articles are in French.

When you show these 4 chords to a person who plays jazz he will tell you that between Fm and Em there is necessarily a downward modulation of a semitone.

In this quite simple example, it is not clear why I play the two initial chords with only one scale: the C major harmonic scale.

What happens when you use this scale on the first two chords is a very special feeling. It's just an example of course. But particular combinations of scales and chords can give very surprising results.

[jancivil]

“the problem of harmony is not as simple and intuitive as you say.”
is for me fortunately

Not to be argumentative, if this is happening for you, more power to ya.
I will address points just to think out loud, and engage you.

<Fm / Em / C7 / C#Maj7

what are the scale you'd use?>

I don’t consider this <one scale has to work> a necessary problem. I think quite differently than that in general.

First problem is supposed to follow a conception of necessary modulation. There is no context for key at this juncture; I don’t know what the goal is harmonically. So modulation to where? No, the first question to a jazzer is what’s the goal.

But let’s go with the scale C D E F G Ab B and your Fm to Em. Looks great!
I would look at those and come up with E F G Ab B (C) immediately. So far, I totally follow. I don’t care about D there one way or another, but we both find that much as possibly essential.

What does “C7” represent here? I, in my jazz thinking, make choices according to harmonic weight or function and to be efficient as becomes necessary. The scale fragment (so far) doesn’t take Bb into account (not that it necessarily must). But your thinking works there, so...

[music-engine]

I'd like to apologize if I used the wrong words, but that was not my intention.

The harmonic sequence was just an example to show how sometimes the attribution of the scale to the chord sequence gives a meaning we would not have imagined before.

An example of that harmonic sequence is here:



But even here :

https://vimeo.com/334719465

I conceived this music starting from this chord/scale association. But obviously I don't impose anything on anyone, and there are a thousand ways of making music, everyone finds his own way.

[vurt]

cant converse the way some of you do...
but listening to it, i do like the composition for industrial disaster, i do feel its let down a little by the instruments here. would love to hear it played by actual musicians
the lead line had me in mind of king crimson, fracture or similar

[jancivil]

“sometimes the attribution of the scale to the chord sequence gives a meaning we would not have imagined before.”

[music-engine]

Thanks vurt !
Maybe it's because I loved King Crimson so much... and it affected my algorithms

[music-engine]

“sometimes the attribution of the scale to the chord sequence gives a meaning we would not have imagined before.”

[counterparty]

“In terms of determining which scale(s) a piece of music is in, I would imagine a probabilistic approach over time.”

Why? What does “over time” do to change the fact of a scale in some music?
Give a musician any tonal music (or modal) and the scale is apparent. It’s not typically more problematic than a determination of key if tonal or harmonic. It’s not a guess.
It’s not quantum theory, a person’s observation or perception changes nothing about what it is.

I agree that it's apparent to musicians; I'm talking about giving it to a computer. All computers have is math.

The point of 'over time' is that notes arrive over time, and the key can change over time.

Imagine that we have notes played on a keyboard. The first note is B: which scale is this in? It's hard to tell. Then we have a second note, then a third, etc. We should be able to work out the scale once we've heard enough notes, but we also have accidentals that aren't actually part of the scale. Although the scale and key are generally apparent to a musician very quickly, as a programmer I need a way to work out mathematically which scale we're using, given the notes that I've seen, the frequency that they appear, etc. There's not really any way to program a computer to 'just know' in the way that a human can (beyond delving into neural networks, etc)

it’s barking up a tree no one has to climb.

We climb trees for fun

[jancivil]

Ok, I still don’t think probability is going to grok human intention for a key change, let alone be a model to generate human-like thought. Musical thought is not terrifically rational methinks
If that’s not the goal...
<There's not really any way to program a computer to 'just know' in the way that a human can (beyond delving into neural networks, etc)>
Right. Until it *knows*, or more primarily thinks, it’s an extension of its programmer. I guess the interest is an interest in problem-solving.

[BertKoor]

Have you ever heard of Lady Lovelace, the daughter of Lord Byron?

[jancivil]

No. I'm familiar with Babbidge's Babbage engine
(where did Babbidge come in)

"The Analytical Engine has no pretensions whatever to originate anything. It can do whatever we know how to order it to perform." - Lovelace

[BertKoor]

She was the first to realise that numbers in a machine could represent musical notes. While that machine never actually worked...

[jancivil]

Ah. Difficult to place myself in a time where that was radical, 'the first', and reading thru the Wiki I'm struck by how relatively novel that still is.
She apparently had high hopes for the thing.

My disposition is that here is a huge gap between a mechanism working a problem essentially, and thought.
I suppose one could be surprised by what an engine does short of having full neural networks and such.

[BertKoor]

It can be done without neural networks. Been there, done that. Well, not with music, but the concept could work.

At work we had fake test data of personal records with random lettered names. Garbage, unpronouncable.

So from the real data I counted how many surnames start with A, B, C etc. How many times is an A followed by an A, B or C etc?
Do that for all combinations of two letters, what is the chance its followed by whatever third.
This way you build up random names on probability, which turn out to be easy to pronounce.

Same thing can be done with musical intervals to build random melodies that are not conplete junk. Usind just a matrix, no network.

Yes, it is far from cognitive...