Network Bending: Hijacking Open-Source Generative Audio AI Models For Creative Purposes

Dive into the network bending, an artistic technique inspired by the established practice of circuit bending but applied to the digital realm of generative artificial intelligence. Technique for artists and enthusiasts from diverse backgrounds to explore the manipulation of open-source artificial neural networks for sound design, live performance, and sound art. The workshop provides an accessible intro to the core principles of generative neural networks, and a fast-forward showcase of the history of neural networks use in arts and music.

Participants are handed technical framework and a set of methods such as latent code manipulation or network rewriting to see firsthand how these techniques can transform existing sounds and generate new ones. A set of pretrained generative audio models trained on Creative Commons material is provided. These models serve as the canvas for the bending experiments, offering a hands-on opportunity for participants to apply learned techniques.

 

*** The workshop is full, registration has been deactivated.


img © Violetta Wakolbinger

Date
10.05.
Start
14
00
End
18
00
Format
Workshop
max. Participants
15
Contributor(s)
Options for Participants

Participants should bring their own laptops with preinstalled and up to date version Pure data (at least 0.54-1).
Basic level of familiarity with Pd is recommended.