Help me to validate my piano startup idea!

[ci4anish]

Dear community!

If anyone ever searched for the notes sheets on the marketplaces, I have an idea how to improve the experience by dynamically changing the difficulty of some parts for the written sheet Here's the link to a screenshot of the quick prototype https://ibb.co/F8G7RQt

I'm imagining my program will be able to configure at least 3 parts: melody line, bassline, and rhythm. This way you can simplify notes if you are a beginner or make them sound more sophisticated.
I would like to hear your thoughts! How useful this feature can be for you? And have you ever had troubles with finding sheets suitable for your playing level?

[anacrusis]

Here are my thoughts:

- I think your startup idea could be of value for people who play various pop/jazz arrangements, but hard to implement for people who play classical music, which I believe is the primary audience on this forum. In classical music, it's not really possible to change the factors you mentioned on a whim in the same piece, because each piece is written exactly as it's supposed to be. There are of course highly simplified arrangements of some popular classical pieces/melodies, but then you would have to create these arrangements of various levels yourself for your startup.

- Which brings me to my next point. You'll have to find someone (or do it yourself) who can notate the various variants of the same piece. If you have three variables you can adjust across five different levels, I guess you'll have to create at least 3x5 = 15 variants of the sheet music for the same piece, which means you'll have to pay someone at least 15 times the cost of making one arrangement for each song on your service.

- Another problem is that each individual variant will have to be engraved which you also have to pay someone to do. Unless you plan on have some tool you'll program to automatically generate the different variants of the sheetmusic, which will not look nearly as good without human oversight and adjustment of the layout. It might be readable, sure, but it won't look as professional as something that is adjusted by hand.

[ci4anish]

Thanks for your thoughts!

Yes, it meant to work for pop/jazz arrangements. It should be a program that will generate those variants so no need to pay. Yes, the challenge is to create a program like this, but I'm wondering if it's something valuable before diving into implementation.

[lostinidlewonder]

Simplifying music is part of a large library of "piano practice" methods that I teach. It is used so that one can practice a piece two hands without problems and then add more notes as what is reduced becomes mastered. This follows the theory that practing in a controlled comfortable manner will require less repetitions to achieve mastery. It is less optimal to play all the notes if you are struggling to play it controlled and merely repeating in that challenging realm until it is mastered, it just takes you more repeats compared to reducing the score first then add. Of course this situation is much more prevalent when you are studying something that is challenging for you, but even with easier works it can be applied and notes added much sooner.

I wonder what kind of algorithm you will use, that will be interesting to see function because the decision making is not always so clear and how to deal with errors will be a large challenge. To present the piece in the multiple difficulty levels you suggested makes your work even more challenging. It would be a very helpful tool if you can automate it accurately.

[ranjit]

Yes, it meant to work for pop/jazz arrangements. It should be a program that will generate those variants so no need to pay. Yes, the challenge is to create a program like this, but I'm wondering if it's something valuable before diving into implementation.

Well, first of all, getting the rights to be able to make a number of arrangements of all kinds of pop songs could be a nightmare. Fake books had been sued in the past, and the landscape is probably worse out there now. Especially, if the software is paid, this is something which you will have to think seriously about. It's not as simple as scraping everything from ultimate-guitar-tabs.

Leaving that aside, you could obviously notate all of the melodies yourself, write down the basslines, and chords, and create rudimentary arrangements of everything given those parameters. I'd say that one problem you might run into is that the arrangements would lack finesse, and users will pick up on the fact that they all sound same-y soon enough.

How do you plan to work around that? Creating a corpus of arranging styles? Some kind of machine learning algorithm to do it for you?

It does sound like a good enough idea tbf.

[dogperson]

I don’t think it would have much usefulness: someone's choice of making the music more difficult might be far different from mine: you might add chords to replace notes. I might change the range or modulate to a different key. It would be a personal choice, I can’t see how a program could provide that number of options.  It would be cost/time prohibitive

If you would be using copyrighted music, you would need to pay royalty fees.  Why do you think sites like music notes charge around $5.99 per download?  That is not all profit.

I would like to add: I find it to be a great concept, nice prototype execution, but just don’t see it working out to be manageable and profitable.

[j_tour]

It is a good idea.

I think there are manifold ways around the copyright question. 

But this is, IMHO, far too ambitious a project to succeed. 

It's not really not a question of me being a naysayer, and it is true that advertisements for proprietary products don't delight me, but here it's an interesting idea.

For the simple reason that there is no deterministic algorithm that can possibly achieve more than the most superficial of results.  And, even those results?  Of what quality?

For any meaningful output, the machine would have had to have been force-fed an enormous amount of material.  And what of the results?

HOWEVER, I do like the idea, and perhaps like some others here and elsewhere, rewriting and reconfiguring scores has been a big part of my own education.

I'd like to see some results, though:  much like the fable of automated automobile driving, nothing would make me happier than to see promising results.

The (i) Why? has been established, the (ii) How? is a mystery, but not important, but it leaves open the question of (iii) What?  And in what way? 

There's just not an ontology of standard notation and the way these notations are interpreted, so it's going to fail in the same ways other automation attempts have.

I wish it were not so, and I don't claim that's a priori true, but I'd like to see a good example in action.  A prototype, if you want.

There could be a value in having a kind of stochastic, John Cage style of "notation terminator" roaming around, ready to transform scores into some variations.

I think that's the best outcome, and I would like to use such a program.