GenAI is a very powerful tool for generating images and other media from a simple textual prompt. Unfortunately, the actual results are very hard to control. Artists do not (1) have an idea of what to paint/sculpt/write, and then (2) simply paint/sculpt/write it. Rather, the process of human creativity is a back and forth between the artist and the medium, until finally settling into a final form.
Is it possible to create an experience where the machine and human are tightly woven into a collaborative cycle pushing each to offer more than either alone? We seek a symbiotic system where the human & GenAI improvise together to discover new forms in the process, while each responding to the other as a work unfolds.
Is it possible to create an experience where the machine and human are tightly woven into a collaborative cycle pushing each to offer more than either alone? We seek a symbiotic system where the human & GenAI improvise together to discover new forms in the process, while each responding to the other as a work unfolds.
Some results
- Use of Low-Rank Adaptation (LoRa) techniques to better control Stable Diffusion models allows artists more agency.
Primary participants
- Yenkai "Kyle" Huang, Nate Branscum, Jackie Guthoni
Advisors: Michael Casey, Michael Cohen, James Mahoney
