Throughout the history of stage and screen, actors have drawn on a variety of techniques to evoke the emotion of the characters portrayed. While systems of learning and practice have varied across schools and traditions, core elements emerge. The acting process, at least within the European Realist context, involves interpreting the script or story and expressing the action and emotion of the character portrayed using body and voice. This approach has moved beyond live theatre and is now dominant in popular entertainment media such as cinema and television. Even in animation, where physical embodiment is transcended and the actor’s body is replaced by the drawn image, similar mechanisms apply. In 3D computer animation, sequences of images are simulated rather than directly drawn, with much of the functionality now automated by digital systems. Recent advances in computer graphics technologies now enable these simulations to be performed in real-time. The implications of this are significant. The research I am therefore undertaking explores the extent to which significant parts of the acting process could be automated. I propose to demonstrate this automation through the use of an autonomously animated computer graphics character: an avatar. The key research question is as follows. Can such an entity enact a convincing, emotionally contagious fictional character portrayal in real-time? In other words, can an avatar act?
To explore this, I am deriving techniques from Konstantin Stanislavsky’s ‘System’ of acting, as well as material drawn from biographical sources. I am also using theories of emotion from philosophy, psychology and contemporary cognitive sciences. These ideas will be developed technically and creatively using procedural programming and Machine Learning techniques, demonstrated within a real-time simulation environment such as the Unreal Engine. Findings will be both discovered and realised using a set of practice research outputs, including screen-based simulative performance, procedural computer animation, and speculative fiction in the form of a written screenplay and its production as a short film sequence.
The research endeavours to shed new light on the polemic surrounding acting technique encapsulated by Diderot’s Paradox, which questions the extent to which an actor has to genuinely ‘feel’ or access the same emotions as the character portrayed. The work will also contribute to the quest for more human-like behaviour within Artificial Intelligence, determining what can be learnt from the very human activity of dramatic acting. The ongoing research in this area is mostly within the field of Affective Computing, which usually attempts to orchestrate ‘real’ emotion within the machine context, rather than simulate ‘acted’ emotion. The work is further situated within the heritage of written fiction, film, animation, theatre and videogames which explore the possibility for a mechanised entity to exhibit human-like emotion. It is also positioned within recent advances in lifelike computer animated digital humans, using speech, gesture and expressions synthesis and response.