CLEAN-VIS is a visualizer for live performance made in max / msp. It recursively collages and distorts small segments of illustrations I did to create morphing abstract textures to the beat of the music. CLEAN-VIS was most recently used at MakeShift Art Space In Bellingham, WA on 9/09.
The device is performable with a series of knobs, in my case I use a MIDIfighter Twister. Two main sliders control fading between predefined parameter states, for easy access while performing. One slider controls parameters relating to image slicing and placement, while the other controls a convolution kernel that is applied to the previous frame every frame. Rather than loading individual presets, or maintaining control of every parameter of the visualiser at once, the sliders allow for continuous fading between a series of presets I have defined. For additional control as needed during performance, the next page of MIDIfighter knobs allows direct access to each parameter of the visualiser.
CLEAN-VIS is being used to create music videos for my forthcoming album.
AI-VIS is a visualizer for live performance made using a combination of python and Max/MSP. AI-VIS plays back videos from a neural network I created and repeatedly trained in python using pytorch. The neural network itself is trained on only one or two images at a time, and operates very simply. The goal is essentially for the network to memorize how to synthesize an input image from a very low dimensional representation (3 x 8 x 8). Upon manipulating that low dimensional representation, distortions in the features of the input image can be made. AI-VIS operates using a series of convolutional layers whose architecture I have not yet settled on. AI-VIS is a part of my own exploration of AI algorithms trained on extremely small training sets.
AI-VIS is being used to create music videos for my forthcoming album.
SLIME-VIS is a visualizer for live performance made using Max/MSP, an iphone camera, glue, borax, mud, among other things. SLIME-VIS applies image distortion algorithms to a series of photos of slime following the peak level of incoming audio. Slime photos were taken by me over the course of several years. SLIME-VIS contains slimes both natural and man-made. No slimes were harmed during the creation of SLIME-VIS. The distortion algorithm remaps pixel placement according to the value of each pixel in auxiliary ‘algorithm’ images I created. The intensity of the remapping follows the peak level of incoming audio.