Search through recordings, the names of events they were recorded at, and the tunes they contain. Currently showing all recordings.

View on Youtube

Jennifer Walshe premiers her new work featuring computer-generated text.

View on SoundCloud

Played by Bob L. Sturm

View on Youtube

I interpret a transcription generated by folk-rnn v2, and analysed here: https://highnoongmt.wordpress.com/2018/01/03/making-sense-of-the-folk-rnn-v2-model-part-5/

View on SoundCloud

A recording of a session featuring Machine Folk music at the 2018 QMUL Festival of Ideas, April 26 2018. Musicians include Bob Sturm (accordion), Sandy Rogers (fiddle), Luca Turchet (mandolin), Emmanouil Benetos (accordion), Michael Mcloughlin (tin whistle), Dan Stowell (melodica and bones), Cornelia Metzig (guitar)

View on SoundCloud

Played by Bob L. Sturm (melodeon) and Carla T. Sturm (flute)

View on Youtube

Played by Bob L. Sturm with the Tip-Top Polka.

View on Youtube

Played by EECSers.

View on Youtube

Played on a C/F Club.

View on SoundCloud

Performed by Bob L. Sturm (melodeon) and Carla T. Sturm (flute).

View on Youtube

Eight short outputs generated by a long short-term memory network with three fully connected hidden layers of 512 units each trained on over 23,000 ABC transcriptions of session music (Irish, English, etc.), and arranged by my own “personal” neural network trained on who knows what for who knows how long (I can’t remember any of the settings) (2015)

  1. Hole's Mill (1:30)
  2. Segue: The Birthday (0:09)
  3. A Fhsoilah Kilnie (1:14)
  4. The Humours Of Time Pigeon (0:33)
  5. The Arian (3:02)
  6. Segue: Larkin's With A Coma Pile Phana (0:28)
  7. Bump Of Howled Sho The fetch (2:38)
  8. Segue: Barch Beach (0:10)

These eight short pieces come from my recent explorations of using deep learning to assist the process of music composition. The training of the text-based network aims to make it produce the “correct” output character following a given input character from training data. The end result is a generative system from which we can sample any amount of output. The network also produces titles in its output. While the system output often exemplifies the conventions in its training data (e.g., Irish and English folk music, see The Endless Traditional Music Session), it sometimes produces surprises. These pieces come from such surprises. (More info: https://highnoongmt.wordpress.com/2015/08/15/deep-learning-for-assisting-the-process-of-music-composition-part-4/)

Acknowledgments: Andrej Karpathy for his open source char-rnn code; João Felipe Santos for help in training and sampling the network; thesession.org contributors.