BachProp | An AI algorithm for music composition
This webpage contains media illustrating the paper published at SMC19 entitled
Learning to Generate Music with BachProp
On each page there is non cherry-picked mp3 renderings of music scores generated by the model in the page title.
The BachProp algorithm has been trained independently on the four original corpora presented in the Original tab.
All other algorithms have been trained on the Bach Chorales corpus only.
Performance of the Ada string quartet at EPFL 2019 diploma ceremony
[0'] Mozart Divertimento
[9'30] BachProp's Nottingham
[10'30] Nordic folk music arranged by the "Danish string quartet"
[14'10] BachProp's Chorales
[16'10] BachProp's John Sankey
[18'30] BachProp's Mozart and Haydn string quartets
[20'] Shine you no more (Nordic folk music)
[24'10] Interview [in french] of Florian Colombo (BachProp designer) and Ken Lila Ashanti (professional musician)
[28'15] The peat dance (Nordic folk music)
The scores of the generated music played at this ceremony are shown below.
© Magistrale 2019 EPFL
Edited score from BachProp trained on the Nottingham dataset
Edited score from BachProp trained on the Chorales dataset
Edited score from BachProp trained on the John Sankey dataset