Show HN: See chords as flags – Visual harmony of top composers on musescore
rawl.rocks127 points by vitaly-pavlenko 5 days ago
127 points by vitaly-pavlenko 5 days ago
I designed a relative piano-roll-based music notation. I used 12 colored arranged in a specific way to make visible the main effects and oppositions of Western tonal harmony. The tonic is always white, so a manual annotation/interpretation is required for each MIDI file.
All chords are flags of three to four colors. Minor mode is darker, major mode is lighter. Colors are arranged in thirds.
I sorted the pieces from simple complex harmony. I also wrote a bit of text to explain what you may see. There's also a corpus of structures: hyperlinks of tags that allow you to find similar patterns throughout my corpus of 3000+ popular pieces.
My method makes chord progressions memorizable and instantly visible in the scores. No preparation of Roman numeral analysis / chord symbols analysis is required. After a bit of training the chords will stare right in your eyes.
It's not synesthesia, it's a missing script for tonal music which makes harmonically identical things look the same (or similar).
I've also recorded lectures on my method in Russian (https://www.youtube.com/playlist?list=PLzQrZe3EemP5pVPYMwBJG...). I'm sorry I haven't yet found time to re-record in English.
I've also sketched a friendlier intro: https://vpavlenko.github.io/d/
Sorry, but this thing won't make any sense if you're color-blind.
It's open-source: https://github.com/vpavlenko/rawl
Earlier context: https://news.ycombinator.com/item?id=39165596
(Back then colors were less logical, and there was no corpus of 3000+ piece annotated yet)
This is one of the cases where choosing a better palette would improve the visualizations As now, there's no relationship between colors beyond different notes, different colors Perhaps choosing similar colors by distance on the circle of fifths or similar They are actually similar by circle of thirds. They go as 1-3-5-7-2-4 over the rainbow. This makes main triads be smooth gradients instead of random three-color flags. 1 and 5 pitches are neutral because: - they are neutral (a hollow tonic power chord) - they don't convey any information about the scale. they only give a reference point to measure everything else against 1-3-5-7-2-4 I would imagine that would be red(1), orange(3), yellow(5), green(7), blue(2), purple(4) But it seems you have white(1), green(3), grey(5), yellow(7), red(2), purple(4) I can't quite see what going over the rainbow in thirds here means, but I can see why a fifth would be neutral. Could you expand upon this? It's rainbow if we skip coloring 1 and 5 and color them with grayscale instead. Here's the question. If we can allow one color not to be "colorful" (chromatic), what pitch would that be? It's the tonic (pitch 1). If we allow two such colors? 5 is a good candidate, it's present in almost all popular scales. (Locrian isn't very popular.) The rest 10 colors go in rainbow by thirds, as you proposed. So, using two grayscale colors, I've reduced the demand to make distinct enough color palette from 12 colors to 10 chromatic + 2 grayscale. Which (10), in my experience fighting with different screens and projectors, is almost the limit of having something stable, distinguishable, nameable and memorable. Funny enough, those 10 can be further split into 5 brighter and 5 darker pitches. Exactly as they turn from bright to dark if we go from bright to dark modes: https://www.dropbox.com/scl/fi/dku7jezn00yltb607j2un/Screens... Pro jazz trombonist here. All there is to usefully say about piano roll notation, with or without colours has pretty much already been said. And if you want more (much more) detail, then Tantacrul (designer on MuseScore) has done a great video. https://youtu.be/Eq3bUFgEcb4?si=lcjA8fF4e3dINvmX The only useful head's up I can give is the current position marker on your playback is quite a long way behind the audio, around 1-2 beats on the one piece I tried. very cool! hookpad/hooktheory/theorytab [1] is a similar idea, but I think the annotations are created using their tool instead of sourced from MuseScore. Yes! hooktheory was my main inspiration over the years. One downside of hooktheory is that it's a reduction which someone should make for you beforehand. That is: - it's losing information - if no one analyzed a song yet, there's nothing you can do about it And, although I don't have an easy way to upload MIDIs yet rather than "you ask me to upload it and I'll do it", I don't do any reduction of the (sonic) score itself. I love that breakdown you did here: https://vpavlenko.github.io/d/ Very cool! Also makes me jump right into strudel.cc and experiment with chords, progressions and melodies. Have you ever explored the idea of shaped notes? There's multiple different approaches with both 4-shape and 7-shape systems being common. But the point is that your color system seems largely correlated to it, and there has been research done on the shape note system. > This simplifies visual analysis: chords like [uncopyable image of color bars]
and other structures become visible, scores become readable and interpretable. The colors are hard-coded to pitches, and so change upon transpositions. For instance a V-I cadence in different keys is functionally the same, but will be colored differently. It does help highlight common tones between nearby chords. Other than that, it's not doing anything for me in terms of seeing function. No, they're not. They're coded to scale degrees. V-I cadence will be the same color in all keys and inversions. OK, so it tracks key changes? How about secondary dominants: V-I cadences targeting any scale degree at any time. I might be using the terminology wrong, but I wouldn't call a I-IV resolution V-I, even though it's the same relative movement. Nor would I call it a key change. I would say downward movement of a fifth is common in western music, but the feel is different for different resolution targets. To me, it seems more clear or explanatory to keep the colors encoded as scale degrees of the current key, not the local cadence-resolution-target or whatnot. In all cases of tonicization I decided whether to notate a brief tonic change or not (piece by piece, case by case): https://rawl.rocks/s/applied/chain_of_dominants You can always manually "probe", minutely "recolor": click on a measure number and hover a potential new tonic. OK, so this is a case of personal knowledge transfer, which is valuable for that reason. You take your expert opinion that certain things are happening functionally in the music, and encode it in color. Why use this piano-roll visualization rather than just color coding notes on sheet music? You lose a lot of other information in the process (like, almost all of it). You're very close to Aikin shape note heads! These help sight-read in any key, since they're always shaped according to whatever the relative major is, and so it's easy to learn the intervals between any two shapes. Yeah this is basically shape notes with colors, in a MIDI format. Not helpful for non-digital musicians Fair point. In my approach I got very dense yet readable information density on the screen. Eg. the entire Mov. 1 of Moonlight sonata is a single screen on my 16" Mac, yet all tonal effects are there, visible: https://rawl.rocks/f/Sonate_No._14_Moonlight_1st_Movement The first movement. Which is 4 pages long. You could display 2 pages on a 16" screen, so you improved density by 2x but you have no articulation, no dynamics or tempo markings, no legato/phrasing notation, no LH/RH indication. If you really wanted to show the harmony as densely as possible you could fit the whole movement on 1 page with figured bass and a comment or two about how to play the arpeggios. I'm not a hater, I encourage the exploration. Just get personally frustrated when we aren't ever building on 1000 years of music notation and instead starting with MIDI and DAW style slop - it's not value-add for serious classically trained artists, and (imo) pushes music dilettantes in the wrong directions So here I stand with dilettantes :) Being a jazz college dropout, I stopped building around standard notation when after two years of training - and I started learning piano at 25 - I still didn't get any fluency in instantly seeing harmonic structure in the scores. I don't care that much about articulation. I can hear articulation anyways. What I care about is the elusive layer of harmony which is so hard to reason about without the right tools. That's why people do Roman numerals, figured bass, all sorts of annotation: https://github.com/vpavlenko/study-music/blob/main/parts/cla... And that's what I was chasing. Something that gives me, a guy who didn't spent formative years of K-12 by sight-reading at the piano, a way to build mental models of how Western harmony works. And here, I believe, I'm with the majority of people. OK let me suggest something to meet you halfway. As someone who is fluent/sight reading/ singing etc and can "hear" harmony and/or "read" it on a page of regular notes (lines and spaces and intervals all you really need, but to your point requires experience), something that would be far more interesting and useful to me, rather than color-coding solfege, would be something more along these lines: Color code each chord by its diatonic value (e.g. in a I chord every note in the triad is red, in a IV chord every note in the triad is blue) and then highlight the extended notes as well (e.g. add 9 is yellow, add 11 is green) That is something that MIGHT be interesting to me (personally - I know I am not your audience). But even thinking it through this technique sort of washes out the interesting bits of substitution/interpretation etc that you can find because you are committing to interpreting a chord with a single root Yeah, I see your point. You want same alterations on different Roman numerals to look the same. (In my current system b9 over I and b9 over V look different.) I briefly explored this appoach:
https://rawl.vercel.app/edit?a=beethoven_op10no1mov1
(Although it was a mess.) In current Rawl terms I call it "annotate modulation at every chord". Back in the days I thought it's better. It turned out that: - You need to do tedious annotation. So you can only work with classical repertoire datasets like https://github.com/MarkGotham/When-in-Rome - Worse even, you need to look at two places simultaneously: to the score and to the string of Roman numerals - And worse still, all chords then look the same. The score itself becomes bland. Most of the time, if you aren't deep inside the jazz repertoire, there isn't that many alterations. So you'd mostly see the same four colors - root, minor third, major third, perfect fifth. - And, as evident in Mozart-Beethoven repertoire, you need to decide what to do with fully diminished chords. Because they sorta don't have a root, acoustically. Thanks for the example. Devils advocate: > You need to do tedious annotation. Worse even, you need to look at two places simultaneously: to the score and to the string of Roman numerals The more work it takes to create your notation, the more useful information it contains. Developer-user tradeoff > And worse still, all chords then look the same. The score itself becomes bland. Most of the time, if you aren't deep inside the jazz repertoire, there isn't that many alterations. So you'd mostly see the same four colors - root, minor third, major third, perfect fifth. Isn't the same true for your current method? Only there are more colors of the rainbow to distract you from how bland it is :) Well said! I agree. I find that standard notation is an amazing tool for conveying how to perform a piece, and absolutely terrible for understanding the harmonic structure and reasoning about a piece. That stuff is all hidden and inferred if you have the years of experience to just know all the intervals present at a glance. Seems to crash Safari on iOS, which is pretty rare for me tbh. Not sure what did there but it could either be profitable or annoying for you.
hamaqueto - 5 days ago
vitaly-pavlenko - 5 days ago
markburns - 5 days ago
vitaly-pavlenko - 4 days ago
midenginedcoupe - 4 days ago
FelipeCortez - 5 days ago
vitaly-pavlenko - 5 days ago
MrGilbert - 5 days ago
coreyp_1 - 4 days ago
kazinator - 4 days ago
recursive - 4 days ago
kazinator - 4 days ago
recursive - 4 days ago
vitaly-pavlenko - 4 days ago
kazinator - 4 days ago
CGMthrowaway - 5 days ago
thrtythreeforty - 5 days ago
CGMthrowaway - 5 days ago
vitaly-pavlenko - 5 days ago
CGMthrowaway - 4 days ago
vitaly-pavlenko - 4 days ago
CGMthrowaway - 4 days ago
vitaly-pavlenko - 4 days ago
CGMthrowaway - 4 days ago
meta-meta - 4 days ago
greggsy - 5 days ago
