The training settings need to be set up before starting this process. FakeApp will use it to store the parameters of the trained neural network. If this is your first time training from person A to person B, you can use an empty folder. You will also need a folder for the model. The training process will convert the face of person A into person B. In reality, the neural network is working in both directions it does not really matter which one you choose as A and which one you choose as B. As a convention, Data A is the folder extracted from the background video, and Data B contains the faces of the person you want to insert into the Data A video. Under Data A and Data B you need to copy the path of the extracted folders.
In FakeApp, you can train your model from the TRAIN tab.
Alternatively, you can attach the videos one after the other using Movie Maker, or an equivalent program. If you have multiple videos of the same person, extract all of them an merge the folders. You’ll then need to run the process twice, to get two folders. Ideally, what you need is a video of person A and a video of person B. The face detection fails fairly often, so expect some manual work to do. Before proceeding to the next step, just make sure that the aligned faces are, indeed, aligned (picture below). You might also see a file called alignments.json, which indicates, for each aligned frame, its original position in the image from which it was extracted.Īfter the extraction process is done, the only thing you need is the extracted folder you can delete all other files. Inside, there will be another folder called extracted which contains the aligned images ready to be used in the training process. If your original video is called movie.mp4, the frames will be extracted in a folder called dataset-video. Clicking on EXTRACT will start the process.
All you need is to specify a link to an mp4 video. Unless you have hundreds of pictures already selected, FakeApp comes with a handy feature that allows to extract all frames from a video. To train your model, FakeApp needs a large datasets of images.
As explained in the first lecture of this course, An Introduction to DeepFakes and Face-Swap Technology, creating a deepfakes requires three steps: extraction, training and creation.