Party Music sounds like glitch music if it were invented in 1941, or a hazy fever dream from Oval’s non-existent Big Band Swing era. Party Music’s Graeme Carr Ellis, a “vaporwave refugee,” has been interested in hypnagogic music, recording dreamy vaporwave as CASE WESTERN. Carr Ellis turned to machine learning and GANs to find inspiration in a new workflow.

This album was produced using SampleRNN, which involves taking whole tracks, slicing them into pieces, and training a model on tones and textures of these fragments of sound rather than a full composition.

The tracks presented here were created from a selection of 1940’s Peggy Lee and Benny Goodman recordings over the course of a few weeks with SampleRNN. The first four tracks are cleaned up from the final raw export, while the last track, “Party Music”, was collaged and treated from an earlier training session.

The result is straight out of the Uncanny Valley Scene: music at once familiar and alien. Fans of the source material may feel like they recognize fragments of voice or the blasts of horns, but the sounds found here are best understood as a skipping CD that never existed. None of the sounds here are “from” the original source material. They are synthesized from an analysis of isolated sound patterns, tones, and textures: a track reduced to pieces, those pieces studied for sonic textures, with new textures extracted and re-assembled through the logic of neural networks.

The output exaggerates the quiet/loud dynamics of big band to an inhuman extreme, stopping completely only to start again to stop completely to start again. In radically transforming the sounds of the 1940s through a post-digital sonic palette, Party Music brings to my mind the distance between early computers of the era and today’s once unimaginable machines.

You can find more from Party Music on Bandcamp: https://partymusic.bandcamp.com/

Produced by Graeme Carr Ellis using SampleRNN.

• Disponible le:
  2022-04-07