<![CDATA[This study aims to enhance the visual quality of deepfake videos by utilizing image recognition technology integrated with DeepFaceLab and DeepFaceLive software. Using an experimental approach with exploratory and quantitative characteristics, the research involves structured stages of deepfake generation, facial feature analysis, and detection evaluation. The Deepfake Detection Challenge (DFDC) dataset serves as the primary training and testing source. The evaluation was conducted using Cosine Similarity, Structural Similarity Index (SSIM), and Peak Signal-to-Noise Ratio (PSNR) to assess visual consistency and quality. The results show high feature similarity and minimal distortion between original and manipulated videos, demonstrating the system's effectiveness in producing realistic and high-quality deepfakes, especially in visually complex contexts such as music videos. Keywords Deepfake, DFDC, face swapping, video synthesis, image recognition, SSIM, PSNR, Cosine Similarity Introduction.]]>
Copyrights © 2025
