The rise of deepfakes has posed significant challenges to the entertainment industry, particularly in Kollywood, where the threat of desifakes (deepfakes) has become increasingly concerning. This paper proposes a comprehensive approach to detecting and mitigating desifakes in Tamil cinema. We present a novel deep learning-based framework that leverages facial landmark detection, convolutional neural networks (CNNs), and recurrent neural networks (RNNs) to identify and classify deepfakes. Our experimental results demonstrate the effectiveness of our approach, achieving a detection accuracy of 95%. Furthermore, we discuss the potential applications of our framework in the film industry and the importance of developing more sophisticated deepfake detection techniques to combat the growing threat of desifakes.