The Infinite Irish Trad Session

My colleague João Felipe Santos and I found some summertime to create:

The Infinite Irish Trad Session

Our interests converged when we both took Andrej Karpathy’s RNN code and applied it to learning to generate ABC music in traditional styles (and here).

So João grabbed all the tunes he could from The Session, I parsed and editted the text file to produce a 13 MB file of 23,962 tunes, and João set to using it to train an RNN with 3 hidden layers having 512 units each. Once finished, João had the system generate a 21 MB text file of new ABC tunes, which I am parsing into its 72,000+ tunes. I convert each ABC tune to MIDI, process it in python (with python-midi) to give a more human-like performance (including some musicians who lack good timing, and a sometimes over-active bodhran player who loves to have the last notes :), and then synthesize the parts with timidity, and finally mix it all together and add effects with sox. Each successful synthesis is then served to my home directory, and a cronjob generates a new playlist every five minutes from the 22000+ tunes already synthesized (that is 18,964,140 kilobytes, which at 128 kbps is a 41 hour trad session!).

Hope you have some craic!

One thought on “The Infinite Irish Trad Session

  1. Pingback: Deep learning for assisting the process of music composition (part 1) | High Noon GMT

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s