a dance-to-divine text generator

by Trisha Chhabra, August Luhrs, and Claire Yuan

made for CIID's Machine Learning for Interaction Design class, July 2017, taught by Andreas Refsgaard and Gene Kogan

Inspired by the Oracle of Delphi, the High Priestess who served as a conduit through which the God Apollo could speak, we created a tool that allows a user to give their god or gods an offering in the form of a dance and receive a personalized message from the divine.

The basic premise is -- a kinect paired with Wekinator tracks the user's dance (mainly just hand height) and will send a single message to a processing sketch that will run torch-rnn on a terminal, seeded by a unique word obtained from the dance. The RNN will then spit out a religious sounding message.

Our RNN was trained on five major religious texts (The KJ Bible, The Quran, The Gita, The Gospel of Buddha, and The Book of Mormon) for around 500 epochs (so, not long enough).

https://twitter.com/oracleofdelphai

https://github.com/DeadAugust/OracleOfDelph-AI