Lamija Halilović
Supervisor(s): Selma Rizvić, Aya Ali Al Zayat
University of Sarajevo
Abstract: With the development of generative artificial intelligence, the distinction between AI-generated content and human reality is becoming increasingly blurred. This research explores the current state of easily accessible generative AI tools by evaluating the visual and vocal authenticity of digital humans. The study features a comparative analysis of eight AI video generation tools, each used to produce an identical video sequence. These videos, along with one real human video, were presented to a diverse group of participants to examine their ability to distinguish between synthetic and real media. Key factors such as facial expressions, lip-sync accuracy, and the presence of the uncanny valley effect were examined throughout the study. The evaluation indicates that even users familiar with artificial intelligence were misled by the high fidelity of avatars, thinking that they were real human beings. Certain AI tools have overcome the uncanny valley, making high-fidelity digital humans natural to the observer. The results mentioned above suggest that the realism of synthetic media has bypassed human sensory perception, highlighting the urgent need for advanced AI verification tools.
Keywords: Human-Computer Interaction, Image Processing, Multimedia
Full text: Year: 2026
