Repository contains experimentation and code related to training AI models for deepfake detection. The code is written to work with UADFV dataset. UADFV dataset is not very big in size (~155MB) and is ideal for running quick experiments. Models trained on the dataset won't be perfect but gives the basic framework of how to train and test the deepfake detection AI networks, e.g, MesoNet or XceptionNet. The experiments provided here are useful to work with other very large datasets: FaceForensics++, DeepFake Detection Challenge (DFDC), Celeb-DF, and DeeperForensics-1.0.