This neural rock album is part of a submission to NIPS 2017 Workshop for Machine Learning, Creativity and Design: "Generating Black Metal and Math Rock"
Music was generated autoregressively with a sample recurrent neural network* trained on raw audio from the album Mirrored by Battles. The machine listened to Mirrored 30 times over several days. The machine generated 1300 minutes of mono audio. A human remixer listened to all of the audio and curated 18 minutes of songs. Titles were generated by a Markov chain. The album cover was generated by neural style transferring the Mirrored cover with a grid of all the Battles album covers.
A human music producer created a stereo enhanced mix of the selected machine generated tracks by "doubling" with two slightly different mono versions. Tracks 1 - 4 were edited in a DAW by snipping multiple clips and cross-fading them together based on the human observed compositional tendencies of the machine.
released January 5, 2018
* SampleRNN [Soroush Mehri et al. 2017]
This site may contain copyrighted material the use of which has not always been specifically authorized by the copyright owner. We are making such material available non-commercially in an effort to educate and advance research in machine learning, generative music, music information retrieval, computational creativity, etc. We believe this constitutes a ‘fair use’ of any such copyrighted material as provided for in section 107 of the US Copyright Law.