A collection of tracks generated automatically.

Learning Music

Watson Beat is a machine deep learning algorithm that learns and creates music. IBM researchers are teaching a complex neural network to understand music theory, structure (pitch, time signature, key signature), emotional intent and co-create music with a human partner. Here are some of the results.

Watson Beat

These are some samples generated by char-rnn when trained on ~14k Irish Folk tunes (downloaded from thesession.org).

Jimmy O'Corlan

Composed by machines. Inspired by soundcloud.com/seaandsailor, learned using github.com/karpathy/char-rnn and synthesized with http://taalmala.com

Read more: http://www.trivedigaurav.com/blog/machines-learn-to-play-tabla/

char-rnn composes a taal!

This track was made using a RNN. It writes the tabs out in ASCII, I imported into GuitarPro, recorded the output, imported that into FL Studio, added some filters and a drum loop and got this. The notes and rhythms themselves are totally unedited.

Recurrence - Music written by a Recurrent Neural Network

The follow up to Recurrent Neural Shady. Char by Char RNN model trained on Jay Z lyrics, then auto-generated lyrics by sampling from the learned distribution (one char at a time).

Jay Z Net

Generated by char-rnn lm_lstm_epoch4.43_2.7623.t7 on all chipzel albums.

Sample.out.2

String Quartet Amped

Stabilize Amped

Juniper Breeze Amped

Fallen Star Amped

Stabilize Original Intrumental

This is the official Soundcloud account for IBM Research. It is managed by Chelsea Dubner and Chris Nay and follows IBM social media guidelines.