Did you ever imagine that you can swap the speech that someone is giving to a speech that someone else has written but the person speaking it is not the original one? This is called as deep fake technology which means that you are faking something without the other person even realising about it. Now, it is found out that you can edit the text from a person’s speech and add words of your own into the speech and it will seem like the person said the words written by you.
This is possible in a talking-head style video where you can add, delete or edit a speaker’s words. Also, this is as simple as editing text in the word processor and no editing or other skills are required for doing this. This new deep fake algorithm will process the audio and video into a new file which will then be matched to whatever the speaker says.
This algorithm is developed by a team from Stanford University, Max Planck Institute for Informatics, Princeton University and Adobe Research. The team says that ideally, this algorithm should reduce the cost of re-shoots because someone said something that should not be there in the video. Or there was something that needs to be added to the shot.
This algorithm requires a training period of just 40 minutes in which it leans the face movements of a speaker and transcript of what is said is also required. This also means that you cannot do this on a short video which might be less than 40 minutes which is generally the case with most of the videos shot casually.
Not only will the algorithm replace the text from the video, but the algorithm will also create a 3D model of the face making the new shapes required. This means that you will feel like the person is actually saying those words and you cannot even guess from their face movements that this is a fake video.