Why ..
I am a musician and a data geek. While working on a DSP project for a VST plugin and it occurred to me that I enjoy the math of music. What better place to explore it than here?
What ..
I spent a few hours recording and mapping out triads on a digital piano. The idea was to create something with AI/ML to use it with .. Maybe a songwriting bot. But surely there are better ideas.
-
This dataset consists of 432 Wav files of piano triads across 6 octaves.
-
There are wavs for each of the 12 major, minor and diminished triad chords in both the root and first inversion positions.
-
The samples are 32 bit 44KHz mono, around 520K in size and 3 seconds in length. The total size of the wavs is around 200MB.
-
The CSV file (triads.csv) contains a list of each of the chord names, their octave on the keyboard, the inversion position, and the notes that make up the chord .
-
The chord names are delimited by two underscore characters which split it into four parts.
-
A lower case 's' following a note name denotes sharp notes/chords. Lower case b indicates flat notes/chords.
-
Cs_maj_2_0 indicates the chord is a C Sharp Major, second octave, root position (inversion).
-
The octaves range from 2-7 .. with octave 5 being "middle C".
-
- Note .. I used E flat instead of D sharp (for all the purists :).
-
The Note1 column indicates the root note and its position. The remaining notes are in Note2 and Note3 columns.
-
All audio samples were recorded on a DAW.
Notebook
Here's a notebook that shows how to load and play the wavs using the chord name and position. It also displays a waveform and the notes that make up the chord.
https://www.kaggle.com/davidbroberts/piano-triads-wavset2/
This notebook is a chord progression maker -> https://www.kaggle.com/davidbroberts/chord-progression-maker