The University of Erlangen-Nuremberg, the Max Planck Institute for computer science and Stanford University are currently working on a software that allows a user to transfer his facial features in real time on the face of another person in a video.
facial features can be transmitted in real time
The developers of the software indicate that the new departure from previous, similar software is that the transmission and the photorealistic neurendern of facial features and their changes and the details on a video works so that the newly-synthesized facial features can no longer be distinguished from a real video.
In fact, the result is amazing when you look at the below linked video. Whether you really can no longer be distinguished from a real video it may be in the eye of the beholder. Sometimes you can see that the facial features seem a bit unrealistic. However, one could at least at first sight and if you is not looking really just get the impression that the person in the video actually owns these facial expressions.
What you can use this software then? For example, for a real-time translation. If one conducts an interview with a person and a translator translates in the background in real time, then you could transfer his face to the person in the video to make it look as if this person speak. Also for virtual-or augmented reality this would be interesting, as a wax figure could be really speak for example in a museum. On the other hand, it is intended also possible to do a lot of Schindlunder with such software. If a person, for example, do not know that it reproduces the features of another person in the video. Still, it’s amazing what is possible today in this regard. / off
No comments:
Post a Comment