K A I R O S is a project of visual poetry empowered by artificial intelligence and datamoshing techniques. All the images were created based on personal photographic material captured through ages and countries. These videos were elaborated with the use of deep neural networks (DNN) applying style transfer. Each style corresponds to a textured or patterned image that is transfered to the content of another image. To apply neural style transfer to videos, we used subsequently a software designed to facilitate animation by projecting a transformed key-frame onto all the other images of the sequence. We hacked this algorithm by feeding it with sequences displaying unknown pictural elements, leading to datamoshing effects. We further used a second datamoshing technique relying on deletion of key-frames and reversed playback (in classical terms). The music was taylormade to accentuate the suggestive effect of our dreamscapes.
With this set of techniques, we aim to reveil the hidden information, the lost memory behind our visual field, through a dynamic of perceptual backpropagation; a process similar to how DNN learn complex natural patterns. These pheno-digital objects (intrinsic oxymorons) manifest our engagement into a dialogue between human & machine pareidolia.
If you want to dream awake, trust the process.
Guillaume Gagnon Lortie