This is a little tool I made that can generate random progressions of 4-part harmony for SATB. It produces notes for 4 voices that are approximately 8-10 notes long. The output is visualized on a graphic staff accompanied by their respective functional chord symbols.
Try out the applet here! Java is required.
The primary reason I built this was mostly because I wanted a way of organizing the harmony knowledge I learned in preparation for the RCM Harmony exam I did this summer. Proper MVC design was fun too. It ended up being a bigger project than I thought. The core part was mapping out the matrices that summarized patterns in the tonic, predominant and dominant harmony sections of a piece of music, which ended up being a central part of my implementation.
I've recently be made aware of HookTheory's API for its crowd-sourced song chord collection.
They contain probabilities of a chord following some previous chord, and of progressions also.
I've written some scripts in Node.js to scrape and analyze.
They are in the
I've also added the data I've collected from its API as a source in the application for generating progressions.
Implementation and Rationale
My idea was to first generate an appropriate functional chord progression. The progression would end in a cadence of some sort. It would then automatically produce a SATB voicing following said progression that would adhere to traditional rules of 4 part writing.
I had brainstormed many ways to potentially do this. I ended up doing it in two ways:
- Chaining through a stochastic matrix, specifically a Markov Chain
- Discrete probabilities would be provided
- Probabilities could give interesting steady-states over the long run
- Numbers would come from published analysis (there are many articles online analysing chord progressions in classical and modern music)
- The hope was that similar patterns would emerge
- But patterns would be simplistic (see Here and Here)
- Using multiple matrices, each of which addresses the Tonic, Predominant, and Dominant harmonies, to chain through
- Wouldn't have probabilities (too subjective)
- Compiled through my own work
- Used adjacency matrices to show paths to and from a chord
- More variations possible of diatonic chord positions (root, inversions), qualities (major, minor, diminished, etc) and types (triads, 7ths, etc)
- Actually had the added effect of being able to generate appropriate passing chords and suspensions, which was a plus
- I designed the matrices with root movement in mind
In the end, I would say that the second way was a lot more successful, in that the music generated made more sense from an analysis point of view.
Rundown of Features
- Can specify key (only major for now) piece is generated in
- Choice between stochastic matrix and tonic-predominant-dominant matrix
- Entry of custom progressions with auto-completion for 4 voicings (try something complicated like
- Able to select which rules of harmony to follow
- Can also identify intervals, make scales in natural and single sharp/flat keys (this needed to exist for the rest to work)
- Export to Lilypond .ly file, which creates pdf score and midi
- Play progression directly through midi sound synthesis
- Tonicization (still need to fill up the matrix though)
- Start and end on tonic
- Soprano within C4 and A5
- Alto within G3 and D5
- Tenor within C3 and G4
- Bass within E2 and D4
- Soprano, alto, tenor cannot be more than one octave apart
- Bass and tenor cannot be more than Perfect 12 apart
- No parallel octaves/unison (toggle-able)
- No parallel fifths (toggle-able)
- No hidden octaves/unison (toggle-able)
- No hidden fifths (toggle-able)
Features in the (Near?) Future
- Minor key
- Modulation (have secondary dominants/sub-dominants, etc, now)
I would be happy if anyone wants to contribute to the the T-P-D matrices.
This project uses the Musica (ver 3.12) font found here. The font provides Unicode support in the 1D100-1D1DD range, which represents the Musical Symbols. The font is offered as "free for any use".
License for HookTheory data is unknown. Please contact HookTheory.com directly.