Banana Vorticism

The method of “Exquisite Corpse” was created by surrealists in the early 1900s as a way of collectively assembling a collection of images or words as a fun way to spend their nights. This game quickly became a way of generating collective art pieces and getting inspiration on the unexpected results of their collective thinking. In the present scenario, artists still use this method to spark creativity.

Though in the present day, the development and use of Machine Learning are advancing at an increasing rate, this technology still has discrepancies and gaps between its analysis and results. The team decided to take advantage of these discrepancies and create Banana Vorticism, a modern version of an Exquisite Corpse that generates unexpected results for sparking inspiration.

Banana Vorticism is a device that allows artists and uninspired creatives to scan anything in front of them, which starts the specific Machine Learning process. This process creates a text of the inputted images and images out of this text. The process runs several times through different machine learning models and provides the result of all of these as well as a final outcome of the process. The device has three modes of usage that loosely determine the type of text and images to be generated.

  • Classic Mode: generate images that resemble more human-like environments or objects
  • Nature Mode: generate more organic outcomes that resemble forms found in nature
  • Bad Trip Mode: generate more disturbing imagery for artists that want to use this as an inspiration

In order to create Banana Vorticism, the team used Runway and a large variety of different machine learning models. To generate the results, they went through a rigorous process of defining what type of feeling they wanted the process results to convey and how to achieve this while combining specific models on the platform. The process also included the linking of all these models and the inputting of an external camera to generate real-time input for the models to start analyzing. Some of the main challenges for the team included dealing with the speed of the linking of these models and working with the form of output for the results.

The team also worked with 3D building technology in order to rapid prototype a form for the device.