The technology behind deepfakes has quickly advanced, utilizing recent breakthroughs in deep learning, especially Generative Adversarial Networks (GANs), to generate extraordinarily synthetic media. A thorough overview of the existing techniques employed in deepfake generation and detection. This includes methods of image and video manipulation and then describes the impacts these have on the social and ethical aspects. The survey incorporates existing methodologies of detection from just obtaining the artifact associated with the use of generating models to applying high-tech deep learning models with limitations and advantages of these approaches. We analyze existing publicly available datasets in detail (Celeb-DF, DFDC, and FaceForensics++) and highlight their strengths and weaknesses as benchmarks for detection algorithms. In addition, the discussion focuses on generalization, robustness, and real-time detection challenges in real applications. It also analyzes emerging solutions like hybrid models utilizing spatial, and temporal features, attack-resistant training framework, and lightweight architecture for real-time applications. We conclude this review highlights the need for more robust, adaptable, and computationally efficient detection methods to combat the evolution of more sophisticated deepfake generation techniques. The aim of this paper is to provide an overview of the current status to guide the path ahead in research and development to counter the malicious use of this technology and to ensure its ethical and secure use.