Textual authoring of animated scenes

Research / Pillars & Groups / Visual Sciences / V-EUPNEA / Internships / Textual authoring of animated scenes
16 January. 2021
As the main goal of this work, we are interested in being able to generate and control the outcome of animated scenes by using textual descriptions. This will provide new intuitive ways to author animations and/or tell stories especially by non-expert users. We get inspiration for this work from a) text based graphic adventure games of the late 80s/early 90s and b) recent work on Deep Learning to synthesize images using textual descriptions. We will investigate some aspects of the domain that include: a) the scale of character control (individual characters, small groups and/or large crowds), b) the appearance of characters (size, gender, race, shape, clothes, etc.), c) character behaviours and interactions, d) environment control (e.g., placement of buildings, roads, shops, characters, fauna, weather control, etc.) and e) camera control (looking through a character’s eyes, panning, zooming, following a character or interaction, etc.).