Abstract
This article develops from the findings of an interdisciplinary research project that has linked film practice research with computer science and law, in an exercise that seeks to digitally resurrect Margaret Thatcher to play herself in a contemporary film drama. The article highlights the imminent spread of machine learning techniques for digital face replacement across fiction content production, with central research questions concerning the ethical and legal issues that arise from the appropriation of the facial image of a deceased person for use in drama.
Highlights
Deepfakes have been used by practitioners since 2017 as a cheap and rapid means of face replacement in digital video
Our work focusses on the creation of synthetized screen characters based on deceased celebrities, developing from an ongoing interdisciplinary practice research project, Virtual Maggie, that explores whether Margaret Thatcher could be digitally resurrected, using machine learning, to play herself in a new period drama3
The ambition of our study was to investigate whether a deepfakes approach to digital face replacement is viable in high definition screen production
Summary
This oppositional subculture initially reveled in its low-grade achievements, with early deepfakes (DFs) produced at very low resolution, typically no more than 294x500 pixels. This article examines the potential development of the deepfakes application of machine learning into mainstream screen production. We predict a near-future in which open source machine learning has democratized access to face replacement in High Definition digital video; we discuss questions around the technological routes to this goal of high quality DFs in fiction production, as well as the major ethical and legal issues that creative practitioners will face when choosing this tool. Our work focusses on the creation of synthetized screen characters based on deceased celebrities, developing from an ongoing interdisciplinary practice research project, Virtual Maggie, that explores whether Margaret Thatcher could be digitally resurrected, using machine learning, to play herself in a new period drama. The current article is both a report on the research findings of Virtual Maggie, as well as a widening consideration of the issues concerned, and is structured around the three themes of that investigation: 1. Creative screen practice: How are the processes, relationships and responsibilities of the filmmaker changed when applying deepfakes to the construction of screen characters?
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
