I’m working on a WebMIDI based wait-for-input piano/keyboard app, like Flowkey, but bring your own music with MusicXML files.

Works amazingly well, using OpenSheetMusicDisplay lib for parsing MusicXML, rendering it to SVG (OSMD uses VexFlow under the covers) & using OSMD to map the music data into my own data model (I track everything I need for timing and visual x,y coordinates inside the SVG).

From there it’s pretty simple, just refactoring my hacky PoC I’ve been using for myself for the last year into something solid based on React + XState.

Anyways, OSMD would let you import arbitrary music as input to this, opening up the possibility of learning specific pieces

That said, it's true that music is not random; learning to read actual music can be a plus over reading an arbitrary sequence of notes (although if you can read random notes fast, you probably also can read non-random music). It could also make sense for music teachers to upload a piece so that their students all work on the exact same music for example.

In that case the app could "flatten" any piece of music into a simple succession of quarter notes. It would not be very different from what it is today, though -- except less random. So, why not.

For the record: I am a professional musician (bassoon player in a symphony orchestra).

Sight reading doesn't really work quite like this when you get better. After a while your brain works more on chunks and patterns. For example, if I see a long scale pattern in F major I don't read very not. I just notice where to go up and down and play. Reading f e d e d c bb ... doesn't work for sight-reading such things in any meaningful tempo.

Having it play along meant you will have to continue even though you make mistakes. Mistakes will happen, and being able to regain ground and go on is important.

But, it's a different topic. Learning (western) music theory involves scales first and foremost. There are hundreds of scales; even if we limit ourselves to 1 major and 3 minor scales (a big limitation), that still leaves us with 12 x 4 = 48 scales. Then you have chord progressions, cadences, types of resolutions, time divisions... rhythm... everything else...

And none of that deals specifically with sight reading. In principle, you could learn music theory without ever knowing how to read music. IIRC, Paul McCartney, arguably one of the best song writers of all time, can't read music fluently (if at all).

I think it takes a very long time (a lifetime?) to actually learn music theory. But it can take from a few months to a couple of years of dedicated practice to learn to read music well. That's what I'm trying to do with this app.

Now, recognizing patterns, chords, chords in arpeggios, chord progressions would be a natural evolution, yes. For now, I don't know how to deal with any of that (esp. from a UI standpoint) but it's something that's on the radar, for sure.

I practice 1-2 hours of scales and chords, tonguing and control every day. Being able to generate weird melodies to challenge all the combinations you have stored in your brain would be amgreat future development. The melodies don't even need to make sense.

One of my favourite etudes are the Deux etudes atonale by Boutry, and some of them are pretty much dice rolls on paper.

Thanks for the follow-up! In your parent comment you were talking about scales and the fact that one is able to guess the next note if they know the scale.

Now random "dice roll" music may be a little easier to produce, and maybe not very far from what we have already.

Can you elaborate on what you would need, specifically? Chromatic random music (using all 12 notes, including accidentals), in the range of the bassoon?

And what do you mean by "play along"? Would you simply need the app to make some sound, or would you expect it to detect and analyze what the player does? (This would require MIDI, and in my experience, the few MIDI wind instruments that exist are unpopular with instrumentalists, to say the least.)

Or -- what about generating random sheet music (PDFs) on demand? That would be a great project, I would love to work on that.

I find the mouse clicking to kinda get in my way, though I see why this makes sense for mobile.

Another cool feature would be if it actually played the tone when pressed. Or chords.

But also super cool as is. I've always had trouble remembering bass cleft notes + I didn't even know all those other clefts even existed.

Sound is in the works. It's a little complicated but it'll eventually work.

MIDI, IDK; I'd like to avoid this being another keyboard playing app. But maybe it's a natural development. We'll see.

But regarding keyboard support, that's a great idea. It would be so much more fluid than using the mouse when on desktop. On it!

Thx!