This is done by means of old childhood photos and video material. Computer software can thus study the mimics and facial features and then put them into a digital model. An actor still has to record text and act as a stand-in, but is “replaced” by the deepfake model in post-production.
“The technology we use is lifelike. We can make one-to-one exact copies of Tom Hanks and the other cast members when they were young,” Umé tells VRT NWS from Bangkok. “The software trained in how a younger version of Tom Hanks used his facial muscles, for example, and then we tried to adapt that to how Hanks is acting now. With our technology, you can no longer tell whether it is a real, young Tom Hanks or a digital version of him.”