DIGIROTO
How can we use generative models to repurpose found footage and generate new compositions and narratives? Can we automate the film-making/animation process with large image databases? What can it reveal about our current zeitgeist, and the way classifying software’s algorithms are built?
An API driven art-making tool that allows users to use Image recognition as a generative video making technique.
Based on a traditional animation technique called rotoscoping (where the original movement of footage is traced/mimicked), DigiRoto is an artmaking tool that allows users to use Image recognition as a generative animation technique.
Using Google’s Cloud Functions such as Cloud Vision API, individual frames of a video are extracted and replaced with a visually similar image from Google Images, resulting in a Rotoscoped composite Animation.
DigiRoto was used on an excerpt from french filmmaker Claude Lelouch’s film Iran (1971), a ‘slice of life’ travelogue film. The resulting output, exported at 2 frames per second, generates new dissected and abstracted compositions and narratives based on the footage, while also quietly revealing algorithmic biases within Image classification systems.