Reactionary: Machine-performed Video Reactions

Overall Concept

Creating a model that generates video of a distilled ‘human reaction’ in order to create absurd relationships between A&B sets of video.

Data

• Youtube Reaction Videos
• Videos that I later make

Questions and Considerations

• What is performed reaction?
• How does the performed reaction relate to the expression of morality.
• How to build a generative adversarial network that accepts two datasets with the understanding that video A results in video B?
• How does machine learning of a tool suggest objectivity?
• How can that suggestion be leveraged in service of the absurd?

Outcomes

Success

• A swirling mesh of human features in an oversaturated day-glo porno-lit youtube-vlogging color palette responding to videos of me handling objects in an untowards fashion.

Failure

• Mainly just technical failure, as in it doesn’t work at all. I am kind of investedness in the ‘incorrectness’ of this gesture so as long as there is some unpredictable relationship between video A and video B with video B hinting at human facial expressions I’ll be happy with it.