The rapid advancement of AI has led to the emergence of sophisticated AI techniques capable of creating fake videos known as deepfakes. These manipulated videos pose significant threats to society, both in social and political contexts, as they can be used for malicious purposes. Deepfakes involve using deep learning systems to produce highly convincing counterfeit videos and digital representations that appear authentic, complete with fabricated images and sounds. Creating deepfakes has become increasingly accessible and straightforward due to advancements in hardware and computing, raising concerns about their potential harm. As a result, there is an urgent need to enhance the identification of such morphed videos. In this paper, we introduced Deepfake_Ensemble (DFE), a strong deep ensemble learning approach to that detects manipulated videos effectively. To construct an improved composite classifier, we suggest integrating various cutting-edge DL classification models. We proved via studies that DFE outperforms other classifiers, obtaining an outstanding accuracy rate of 99.15% in identifying deepfakes. These outstanding results indicate the superiority of our method and lay a strong foundation for building a real-time deepfake detector. Given the prevalence and potential harm caused by deepfakes, our research aims to contribute significantly to the ongoing efforts to combat their dissemination and protect society from the malicious use of this technology.