Publication > community > Tune Out

Tune Out

This collection is an exploration into new forms of art. Combining original music with machine learning art to produce something completely unique.

For the music I wanted to explore many different sonic textures and tones. From simplistic soundscapes to hints of full classical compositions, all condensed into sixteen second loops. All the music was written and performed by myself. I then took the music to be mixed in a studio and mastered professionally so that I could provide the best sonic experience possible.


This is my first major step creating with the new art form I like to call Machine Learning Art, also known as AI assisted art. The basis for each of these first six channels is a GAN interpolation video. This is a journey between predefined points inside a GAN model, getting just a taste of all the potential generative possibilities of these magical algorithms. Channels 1-5 were made with existing models, while Channel 6 was made from a personally trained model using data/images from the artists GDM. From there I send the videos through a StyleTransfer algorithm to give them a little extra texture and life.

Once the initial visuals and music are complete I frame them in a modified TV image and throw in some AfterEffects magic (are the walls moving?) to give the collection a sense of cohesion.

Without further ado, please enjoy the first half of my collection, Tune Out.

Attributions:

Mixing: Cameron Spies from Night Heron - https://twitter.com/nightheronmusic

Mastering: Hobbes - https://hobbess.com/

GAN model used in CH1-CH5: Joel Simon's artbreeder - https://www.artbreeder.com/about

Dataset / images used to train CH6's model: GDM - https://twitter.com/GriffinDart

TV image base - Adobe Stock standard license