3.1 Video Artifact Production
The study roughly took the form of an 2x2 test of four synonymous
videos. The videos were identical save for their spokesperson –
male/female and organic/synthetic.
Two of the videos were created using organic, human spokespersons. Two
professional actors were hired to pitch an investment opportunity for a
fictitious entrepreneur-owned plant store, Plant Haul, seeking to expand
beyond its initial location. One actor was male and the other female,
but they were otherwise cast to be within the same general demographic
categories (age, ethnicity, general bodily ability, and native spoken
language). Filming was completed in a professional studio with a neutral
backdrop and with the assistance of camera and sound technicians. The
actors spoke from a teleprompter with a script written to be generally
persuasive but not overwhelmingly so, so as to allow room for the
spokesperson to matter. The actors were given wardrobe to match the
wardrobe of the Synthesia avatars and were directed to restrict hand
gestures, given that the avatars could not do so at the time of filming.
Clips from the actors’ best takes were edited in with B-roll and titles
(on-screen text), in line with modern professional video conventions via
professional editing software.
Two videos were then created using Synthesia’s synthetic, AI-powered
avatar program. The avatars were chosen to match the general
demographics (sex, age, general bodily ability, and ethnicity) and
American English voices were chosen from those provided in the Synthesia
system. The same script that the organic spokespersons used was inputted
into system, which generated both the conversational voice performance
and animated the mouth and face of the humanlike avatar. The B-roll and
text were edited in within Synthesia’s editing platform to be identical
to the organic videos.