Disney has patented an AI that can rejuvenate or age actors with a click
In just five seconds, FRAN is able to change the appearance of a face, rejuvenating or aging it as needed.
Walt Disney ‘s research and development team has created an artificial intelligence capable of automating the aging and rejuvenation process of actors. This way, Disney’s production companies can use the same actor to shoot scenes set in different eras.
The AI is called FRAN . Obviously, however effective and fast the technology devised by Disney is, it is still necessary to manually retouch the images generated by the AI, in order to obtain a cleaner and more satisfactory result. The AI – explains Engadget – takes just five seconds for a single frame.
Changing an actor’s age has always been an expensive effect, and in some cases very difficult to achieve with satisfactory results. Especially when it is necessary to rejuvenate an actor who is now of his age – think of the emblematic case of The Irishman -, the risk of involuntarily resulting in the grotesque is always very high. The AI developed by Disney makes it possible to avoid long (and expensive) make-up sessions, and the results look extremely promising.
The Disney developers explain that previous software created for this purpose were typically quite effective with static images, while they gave more than a few problems with moving images, ie with videos. “This is, in our view, the first truly production-ready software that offers a highly practical and fully automatic solution for changing an actor’s age in a video image,” say the researchers. who participated in the development of AI.
The Disney researchers explained that the neural network underlying the FRAN technology was not trained using images of real people: “it would have been impossible, because photos of the same person, in the same pose, with the same light, but in times of his different life”. So they came up with a more ingenious trick: They used another neural network to generate millions of images of non-existent people. The same fictitious person was then generated at different ages, thus creating a huge dataset.
By feeding all this information to the neural network, they managed to create a system capable of understanding how to age or rejuvenate a real person in a completely automatic way.
