Abstract
Current state of computer graphics method allows for visual fidelity of computer-generated characters (CGC), with accurate human non-verbal behavior, including gesticulations and facial expressions. Utilizing CGCs in sex research offers many advantages (e.g., customization, financial costs, ethical and legal issues), but can also be considered risky.
According to the Uncanny Valley hypothesis, the realistic CGC often appear unintentionally eerie or creepy due to subtle imperfections that may be unsettling for a viewer. Increased human likeness of a CGC, increases perceived affinity of CGC, but only until a certain point.
The sharp decline of the affinity occurs, when CGC looks almost exactly as human. The pronunciation of this effect is even sharper, when CGC is moving.
People are sensitive to motion patterns and imperfect or faulty animation of motion patterns may contribute to overall impression of eeriness. The aim of this pilot study was to test the differences in ratings of participants during exposure to stimuli depicting women with different level of realism and motion.
As stimuli we used five female CGCs performing three moves (walk, dance, swirl), presented in randomized order (5s), together with stimuli depicting five real women (selected from streaming sites such as Pornhub to match for looks and behavior of CGC) of a same length. After videos, still picture of each character from the frontal view was presented.
Participants rated the realism, eeriness, and attractiveness of women depicted in the stimuli using 9 semantic scales, three for each dimension. For this study, the set of naked female CGC we created in collaboration with graphic studio 3dsense.
Movements of CGCs were animated based on real adult actress, recorded using a network of 20 Raptor-4 cameras and remapped to a virtual character. For facial expressions, actress's face was recorded using Faceshift software solution in combination with an Intel RealSense depth camera.
Movements and facial expressions were combined using the Autodesk Motion builder software solution. Five attractive female CGC with sufficient variability (hair color, body shape) were purchased at DAZ Studio.
We hypothesized that our set of CGCs - visually highly realistic with animated motion patterns based of real people - would go beyond the Uncanny Valley. They would not evoke feelings of eeriness and ratings of participant would not differ much based on level of realism and motion.
If that is the case, created set of CGC could be used as a stimulus material in sex research. Preliminary results together with a set of highly realistic female CGC will be presented.