Computational methods for creative inspiration in thematic typography and dance

Abstract

As progress in technology continues, there is a need to adapt and upscale tools used in artistic and creative processes. This can either take the form of generative tools which can provide inspiration to artists, human-AI co-creative tools or tools that can understand and automate time-consuming labor so that artists can focus on the creative side of their art. This thesis aims to address two of these challenges: generating tools for inspiration and automating labor-intensive, tedious work. We approach this by attempting to create interesting art by combining the best of what humans are naturally good at -- heuristics of `good` art that an audience might find appealing – and what machines are good at – optimizing well-defined objective functions. Specifically, we introduce two tasks -- 1) artistic typography given an input word and theme, and 2) dance generation given any input music. We evaluate our approaches on both these tasks and show that humans find the results generated by our approaches more creative compared to meaningful baselines. The comments received from participants in our studies reveal that they found our tasks fun and intriguing. This further motivates us to push research towards using technology for creative applications.