Skip to content
Gatekeeper Radio
Streaming Now:

    Deep dreams and analog zombies

    • Electronica
    • Experimental
    Ephraim Wegner, Thomas Wenk and Daniel Bisig combine digital and analogue technologies in installations, compositions and collaborative improvisations. In the radio show "Deep Dreams and Analog Zombies" they present some of their recent works and talk about the technical and compositional principles involved. E. Wegner will present two compositions "Acuter Swim" & "A Piece of Music for Human and Object Movements" using a digital instrument that combines a physical model of a sounding sphere with the playback of soundfiles using a granular synthesis approach. The instrument is being developed for IDM (Ideosyncratic Dance Machine) and draws heavily on the previous work called Puppeteering Al. It is partially controlled by body movements based on motion-capturing recordings by Marcella Centenero. Both the pre-recorded audio and the motion-capturing data are reorganised into clusters that describe different qualities that emerge from the original materials. The audio clusters relate to the spatial movement trajectories by mapping their characteristics onto the coordinates of a sphere surrounding the virtual dancer. The next piece "TAPE DECLARATION" is by T. Wenk. Focusing on the sounds of cassette recorders, such as pressing the buttons, rewinding or fast-forwarding a tape, or using the play function in combination with different playback speeds, he will generate a live piece of music or non-music. The final improvisation is a live presentation of the previously released title "Analog Zombies in Deep Dreams" (DEGEM CD 23), in which the artists build a bridge from analogue media of the last century to the current possibilities of machine learning. For this purpose, the technical sounds of cassette recorders were processed by a machine learning-based approach called Deep Dream. Deep Dream audio is an adaption by D. Bisig and E. Wegner of the Deep Dream image processing method. The method employs a neural network that has been trained to classify the operating and running noises of cassette recorders. This network can be used to create new sound material. In an initially noisy signal, the acoustic properties are iteratively amplified, to which selected network layers respond with high activity. Image: Adrian Schwarz, Reenactment of Music for a Long Thin Wire

    Full show Description

    Ephraim Wegner, Thomas Wenk and Daniel Bisig combine digital and analogue technologies in installations, compositions and collaborative improvisations. In the radio show "Deep Dreams and Analog Zombies" they present some of their recent works and talk about the technical and compositional principles involved. E. Wegner will present two compositions "Acuter Swim" & "A Piece of Music for Human and Object Movements" using a digital instrument that combines a physical model of a sounding sphere with the playback of soundfiles using a granular synthesis approach. The instrument is being developed for IDM (Ideosyncratic Dance Machine) and draws heavily on the previous work called Puppeteering Al. It is partially controlled by body movements based on motion-capturing recordings by Marcella Centenero. Both the pre-recorded audio and the motion-capturing data are reorganised into clusters that describe different qualities that emerge from the original materials. The audio clusters relate to the spatial movement trajectories by mapping their characteristics onto the coordinates of a sphere surrounding the virtual dancer. The next piece "TAPE DECLARATION" is by T. Wenk. Focusing on the sounds of cassette recorders, such as pressing the buttons, rewinding or fast-forwarding a tape, or using the play function in combination with different playback speeds, he will generate a live piece of music or non-music. The final improvisation is a live presentation of the previously released title "Analog Zombies in Deep Dreams" (DEGEM CD 23), in which the artists build a bridge from analogue media of the last century to the current possibilities of machine learning. For this purpose, the technical sounds of cassette recorders were processed by a machine learning-based approach called Deep Dream. Deep Dream audio is an adaption by D. Bisig and E. Wegner of the Deep Dream image processing method. The method employs a neural network that has been trained to classify the operating and running noises of cassette recorders. This network can be used to create new sound material. In an initially noisy signal, the acoustic properties are iteratively amplified, to which selected network layers respond with high activity. Image: Adrian Schwarz, Reenactment of Music for a Long Thin Wire