AUDIOSTELLAR
MACHINE LEARNING & ART LAB
A project by: Mariano Sardón (Artist - UNTREF) and Bruno Mesz
Coordinator: Leandro Garber (UNTREF)
Research Student: Tomás Ciccola (UNTREF)
Our team focuses on the creation of software tools for experimentation, understanding and generation of images and sound. AudioStellar is a sampler-like musical instrument for latent sound structure discovery and experimentation. It processes a user selected folder containing audio files and generates an intelligent sound map placing each audio file as a point in a 2D space.
Nearby points correspond to spectrally similar sounds (i.e similar timbre) while far away points are dissimilar ones. This allows the sound artist to explore a sound library or field recording aided by machine intelligence that reveals a latent structure present in the input sound files. AudioStellar features multiple modes allowing novel composition and sound design workflows.